Introduction

Dental calculus, or mineralized dental plaque, has become an important resource for understanding the evolution of oral microbes1. Dental calculus forms consistently throughout an individual’s lifetime on the supra- or sub-gingival surfaces of the teeth2,3,4. The calcium phosphate mineral structure persists in archaeological contexts, preserving biomolecules such as DNA, proteins, metabolites, and microfossils5,6. The preservation of dental calculus in the archaeological record further allows for a geo-spatial and temporal assessment of microbial variation in the human oral cavity.

Previous studies have highlighted differences in oral microbial composition and abundance in relation to diet7,8 and the environment9,10 in humans, as well as differences in oral micriobiota between humans and non-human primates11,12,13. Recent research has also demonstrated that the oral microbial communities identified in human dental calculus samples are largely similar and conserved through time14,15,16,17. Yet, individual microbes show evidence of strain replacement and lineage diversification associated with human migrations17,18,19,20. Notably, lineages of oral microbes such as Tannerella forsythia associated with pre-European contact American populations have been identified, with these differing from those found in post-contact, historic, and modern American populations18,21.

Investigating how past human migrations have shaped oral microbiome composition and diversity is important for understanding their potential impact on present-day oral health disparities. One of the largest population movements in recorded history involved the forced enslavement and transportation of over 15 million persons from the African continent across the Atlantic Ocean22. The Trans-Atlantic Slave Trade fed the system of extractive colonialism driven by European imperialist desires for dominance and wealth, as well as religious and racial superiority23,24,25. These practices resulted in over 300,000 African persons being forcibly brought to eastern North America between the 16th and 18th centuries alone23,26,27. Enslaved persons endured horrendous conditions in the Middle Passage, and many thousands of lives were lost during their transportation to the Americas27. Individuals sold to European-American enslavers were forced to perform domestic and manual labor, sowing the economic and social foundations of colonial America on the basis of race28. Within this history, evidence of daily resistance in enslavement conditions have been recorded in first-person oral and written accounts29,30,31, as well as demonstrated in mortuary practices32 and foodways, including diet and food preparation techniques33.

Examining the lived history of enslaved persons in direct conversation with descendant communities and/or other racial, cultural, geographic, and spiritual community stakeholders is of paramount importance34. In parallel with movements in Indigenous data sovereignty and bioethics35,36, these discussions have shaped contemporary conversations surrounding consent and the role of community engagement in ancient DNA (aDNA)37,38,39 and microbiome studies40,41,42. Such discussions help to foster scientific accountability by making projects attuned to the priorities of impacted communities, particularly those of descendant community members. When done in collaboration, paleogenomics and microbiome research can offer crucial insights into the histories and lived experiences of subaltern communities, including aspects such as geographic origins, biological relationships, diet, and disease, which are often not well documented.

The Anson Street African Burial Ground (ASABG) Project draws from these theoretical frameworks to investigate the lives of 36 individuals of African descent that were found in downtown Charleston, South Carolina, USA. Dr. Ade Ofunniyin and the Gullah Society, Inc. (now the ASABG Project) set up a community-based initiative to understand the history of these 36 Ancestors (collectively referred to as ‘The 36’ or ‘The Anson Street Ancestors’; the word ‘Ancestors’ was chosen by the community as an honorary signifier of their connection with the living). Over the course of the ASABG Project, we held a series of Community Conversation events, conducted educational outreach with local schools and universities, and created opportunities for arts programming to create space for expression and reflection43. Persons who attended our Community Conversation events wanted to understand the demographic profiles, relatedness, lived experiences, dietary habits, and health of the Ancestors. These activities motivated the creation of a Naming Ceremony, during which the 36 Ancestors were bestowed honorary names, instead of burial numbers, which are used in this text (see Fleskes et al. 44. and Gilmore et al. 43. for more details). The Ancestors were reburied in May 2019 on the grounds of their original resting place during a Reinterment Ceremony as an act of celebration and remembrance.

Thus, with community support, we undertook a multi-method investigation combining aDNA, isotopic, archaeological, and archival research. Our research indicated the individuals had primarily African genomic ancestry and were interred between 1760-1790 C.E (Supplementary Note 1)44,45. Chromosomal sex was assigned based on the presence of XY or XX sex chromosome data, revealing that both women and men were present in the burial ground45.

Here, we report the findings from our study of the oral microbiomes of the Anson Street Ancestors using an aDNA and shotgun metagenomics approach grounded within a community engagement framework. We evaluated the preservation and authenticity of ancient oral microbial communities and characterized the relationship between dental pathologies and oral microbiome profiles. We also assessed differences in oral microbiome composition of the Ancestors and other ancient and modern populations using publicly available oral metagenomic data. In addition, we reconstructed the genomes of two oral bacteria, Tannerella forsythia and Pseudoramibacter alactolyticus, to investigate patterns of strain diversity and replacement in relationship to large-scale forced population migrations of the Trans-Atlantic Slave Trade and European colonialism. We further investigated dietary intake through functional profiling of metabolic pathways associated with different diets. Overall, our findings provide important insights into the oral health, makeup of oral microbial communities, and the evolution of oral taxa in the context of the African Diaspora.

Results and discussion

Presence of dental calculus and authentication of historic oral microbiomes

Dental calculus deposits were obtained from the teeth of 14 Ancestors, which included incisors, canines, premolars, and molars (Table 1; Supplementary Fig. 1; Supplementary Data 1). In cases where a limited amount of calculus was available, deposits from multiple teeth were collected.

Table 1 Demographic profiles and tooth sampling of the Anson Street Ancestors
Full size table

We were able to successfully generate double-stranded DNA libraries partially treated with uracil-DNA-glycosylase (UDG) from dental calculus samples of 12 Ancestors. DNA libraries were shotgun-sequenced, generating between 17,090,464 to 51,541,395 raw reads (Supplementary Data 1). More than 80% of reads were kept following trimming, quality-filtering, and merging, resulting in 13,773,352 to 49,449,332 reads.

To estimate the endogenous human DNA content in the dental calculus samples, the processed reads were mapped to the human reference genome (hg19) (Supplementary Data 2). This resulted in 5067 to 58,170 unique mapped reads, with average fragment lengths between 56-92 base pairs (bp). Extraction and library negative controls contained 33 to 1152 unique reads mapping to the human genome. In all, the percentage of endogenous human DNA content in the Ancestors’ samples ranged from 0.02–0.16%.

Human DNA damage patterns were assessed using MapDamage46. The majority of the Ancestors’ samples had 2.72–7.21% of 5’ terminal nucleotides with C-to-T misincorporations and 2.65–7.96% of 3’ terminal nucleotides with G-to-A misincorporations (Supplementary Data 2; Supplementary Fig. 2). Samples from three Ancestors (Zimbu [CHS13], Nina [CHS20], and Kwebena [CHS21]) did not show damage patterns indicative of aDNA, although the mean fragment lengths of the human DNA were in the expected size range (<100 bp). Due to the possibility of low-level contamination from modern sources, we excluded data from these individuals from statistical comparisons of microbial taxonomic and functional profiles. Prior to further analyses, reads mapping to the human genome were removed from the Ancestors’ metagenomes, resulting in 13,734,044 to 49,410,798 reads across all samples, hereafter referred to as ‘analysis-ready’ reads.

We conducted microbial species-level taxonomic profiling by screening the analysis-ready reads using MetaPhlAn447 and estimated the proportion of potential contaminants through SourceTracker48 analysis. Modern calculus, plaque, feces, skin, and soil samples were used as ‘sources’. For all 12 Ancestors, the majority of reads present in the libraries were attributed to microbes found in oral sources (Fig. 1; Supplementary Data 3).

Fig. 1: The Anson Street African Burial Ground site displaying burial numbering and dental calculus information.
figure 1

The Ancestors from whom dental calculus was sampled and sequenced are highlighted in red, and those where dental calculus was sampled but were not sequenced due to low recovery are highlighted in blue. The Ancestors with no dental calculus are denoted in grey. The tooth or teeth with dental calculus deposits that were collected from each Ancestor are shown. Pie charts display the results from the SourceTracker analysis. The original site map was published in Fleskes et al. 44.

Full size image

Characterization of oral microbial profiles

A total of 239 microbial species were identified in the Ancestors’ calculus samples using MetaPhlAn4 (Supplementary Data 4–5). Most of these species are characteristically found in mature oral biofilms such as dental calculus16. While no microbial taxa were detected in the laboratory controls, we filtered out taxa whose origin could not be authenticated as being oral, as well as those found in only one sample prior to downstream analyses (Supplementary Data 6). After filtering, we visualized the 25 most abundant species using a heatmap (Fig. 2A).

Fig. 2: MetaPhlAn4 results.
figure 2

A A heatmap showing the 25 most abundant species. B The histogram indicates total number of taxa identified for each Ancestor.

Full size image

The most abundant oral bacterial species identified was an unclassified Firmicutes bacterium (SGB71327), followed by Methanobrevibacter oralis and Actinomyces dentalis. The presence of a high number of unclassified Firmicutes species highlights the need for deeper sequencing of oral taxa and refinement of genomic databases to better characterize oral microbiomes. We also detected taxa known to be early colonizers of oral biofilms49,50,51, such as Actinomyces and Streptococcus47,49,50. Bridging species, such as Fusobacterium nucleatum, which represent a transition from the primarily aerobic early colonizers to facultative anaerobic late colonizers, were also identified51,52,53. Late colonizers, including Tannerella and Porphyromonas species and Parvimonas micra, were found in most of the Ancestors49,52,53. Known opportunistic pathogens such as Treponema denticola, Tannerella forsythia, and Porphyromonas gingivalis, which are collectively known as the ‘Red Complex’ and associated with periodontal disease54 were also detected. The unique MetaPhlAn4 profile observed in Zimbu [CHS13] resulted from the lack of identifiable taxa (Fig. 2B).

Correlation between oral health and oral microbial communities

To understand the relationship between the oral metagenomic profiles and oral pathologies, we first examined the Ancestors’ complete dental arcade and the individual tooth and/or teeth sampled for dental calculus (Table 2). Carious lesions were the most commonly identified, with 10 Ancestors displaying one or more lesions. Other pathological changes included those associated with oral health and infection (abscess, antemortem tooth loss, caries, infection, and periodontitis), wear (generalized, occlusal attrition, pipe facet, and tooth modification), enamel defects (enamel hypoplasia and staining), and supra-eruption and tooth fractures. These pathologies were also previously observed in the Ancestors lacking dental calculus, except for malocclusion44.

Table 2 Dental pathologies and other notable dental features observed in the 14 Anson Street Ancestors analyzed in this study
Full size table

When analyzing the complete dental arcade, we found that the Ancestors with dental calculus deposits either displayed very few or multiple oral pathologies. Individuals with multiple pathologies were mostly associated with oral health or generalized wear (Table 2; Supplementary Fig. 3A). Analysis of the individual teeth with dental calculus deposits indicated that six Ancestors had dental pathologies, classified as either dental caries or occlusal wear (Table 2; Supplementary Fig. 3B). These observations align with previous conclusions from a complete assessment of the 36 Ancestors, suggesting they had overall poor oral health44.

Next, we conducted principal component analysis (PCA) to assess correlations between dental pathology metadata variables and the oral microbiome composition of the Ancestors (Supplementary Data 6). Bi-plots displaying the PCA loadings (Fig. 3A) were generated to illustrate how different variables affect sample distribution in the PCA space55. For the Ancestors, negative PCA loadings were primarily influenced by unclassified Firmicutes species and Methanobrevibater oralis (Fig. 3B). Positive PCA loadings were characterized by anaerobic or facultative anaerobic taxa, and driven by the presence of Actinomyces dentalis, Olsenella, and Pseudopropionibacterium (Fig. 3B).

Fig. 3: Bi-Plot analysis of the MetaPhlAn4 mean relative abundances for the Anson Street Ancestors.
figure 3

A Bi-Plot analysis, with black arrows representing the top 10 species with the highest PCA loadings, and grey arrows representing the top 10 species with negative PCA loadings. B The table of species corresponding to the arrows, matching the arrow numbers.

Full size image

Subsequently, we examined whether there was a relationship between the number or type of oral pathologies and the oral microbiome profiles. We integrated the comparative oral pathology metadata with the previously generated PCA distributions and conducted PERMANOVA association testing to assess their significance (Supplementary Fig. 4; Supplementary Table 1). None of the dental pathology metadata variables were found to be significant. However, no Ancestors with dental calculus deposits contained a complete dental arcade, thereby limiting the completeness of this assessment.

Following this methodological approach, we investigated how migration might affect oral metagenomic composition by integrating previously published strontium isotopic data indicating the likely place of childhood residence44. Of the authenticated oral metagenomic profiles from the Anson Street Ancestors, only Banza [CHS01], a young male, likely resided in West or West Central Africa during his childhood and lived some of his adult life in North America. However, PERMANOVA testing between PCA distributions and residency patterns did not show any significant associations (Supplementary Fig. 4A; Supplementary Table 1). Thus, additional data are needed to determine how residency may affect oral microbiome composition throughout a person’s lifetime.

These findings contribute to the ongoing discussion about the links between oral health and microbial community structure in dental calculus14,15,18,56. We observed numerous taxa associated with oral infection and pathogenicity in the dental calculus of Ancestors with poor oral health, consistent with previous studies56. However, our results did not reveal a direct relationship between specific microbial community profiles and oral pathology, as seen in other studies15,18. Together, these results highlight the limitations of using microbial community profiles of archaeological dental calculus to directly estimate oral health of past populations.

Comparative analysis of dental calculus microbiome composition

To compare the oral microbial composition of the Anson Street Ancestors with other dental calculus samples, we selected comparative data based on geographic location, time period, sample number, and availability and conducted a PCA using the MetaPhlAn4-based relative abundances of detected species (Supplementary Data 7–8). The dental calculus samples from the Ancestors grouped on the right side of the plot with positive PC1 values but displayed variability along the second PC dimension (Fig. 4A). Notably, they aligned with other archaeological individuals from the historic period in Asia (Japan) and Africa (Morocco, South Africa) along the positive PC1 axis. Conversely, modern European samples (Spain) exhibited negative PC1 values to the left on the PCA plot. PERMANOVA analysis confirmed significant differences based on site, time period, and location (Supplementary Table 2).

Fig. 4: Comparative PCA analysis of the Metaphalan4 mean relative abundances at the species level.
figure 4

A PCA depicting the Anson Street Ancestors and comparative reference populations. The legend and color symbols are denoted on the bottom of the plot and are additionally shared with subpanel C. B A PCA analysis, with coloring representing the number of identified species. Legend symbols are denoted in the same subpanel. C A Bi-Plot analysis, with arrows denoting the top 10 species with the highest and lowest PCA loadings shown in the corresponding table. Legend and color symbols are shown above the plot.

Full size image

To investigate whether the PCA distributions were influenced by the number of identified species, we colored PCA points by species number (Fig. 4B; Supplementary Data 9). Our results revealed that dental calculus samples with fewer identified species appeared to cluster along the positive PC1 axis. By contrast, modern dental calculus samples had a relatively higher number of identified species and were positioned away from archaeological samples. Samples with very few identified species (< 25) clustered in the middle of the PCA. PERMANOVA analysis suggested this pattern was not significant (Supplementary Table 2). Several factors, including DNA preservation and the lack of appropriate reference genomes in databases, may contribute to fewer species being identified in certain ancient dental calculus samples15.

We conducted a bi-plot analysis to examine the impact of species assemblages on PCA distributions, focusing on the top ten taxa influencing positive and negative loadings for the first dimension (Fig. 4C). The species influencing positive loadings in this analysis were observed in a previous bi-plot featuring only the Anson Street Ancestors (Fig. 3). In contrast, the majority of species contributing to negative PC1 dimensions were rarely detected in the Anson Street Ancestors but more commonly found in modern dental calculus samples. These taxa, such as Ottowia, Neisseria sicca, and others, represent a spectrum of both early and late colonizer species commonly encountered in the oral cavity50,57,58. Differences in species composition between the historic and modern dental calculus samples may also be influenced by modern oral hygiene practices, which can prevent dental plaque buildup and result in higher levels of early colonizer taxa1, or a number of other factors related or unrelated to host ecology15.

Genome reconstruction of oral bacteria, Tannerella forsythia and Pseudoramibacter alactolyticus

A total of 173 taxa identified using MetaPhlAn4 (Supplementary Data 10) were targeted for microbial genome reconstruction using a reference-based mapping approach18. We were able to authenticate the recovery of partial or near-complete ancient genomes for several species of interest, including Tannerella forsythia, Treponema denticola, Streptococcus constellatus, and Pseudoramibacter alactolyticus (Supplementary Fig. 5A–D). Excluding strains that were recovered at low-coverage or those with high heterozygosity values indicative of the presence of multiple strains within a dental calculus sample restricted our phylogenetic analyses to two species, T. forsythia and P. alactolyticus. The former is an opportunistic oral pathogen long-associated with periodontal disease, whereas the latter is an understudied oral microbe commonly found in some ancient dental calculus samples and is a primary endodontic pathogen in modern contexts56,59. We were able to successfully reconstruct and authenticate a partial T. forsythia strain from Tima [CHS31] and two partial P. alactolyticus strains from Lima [CHS03] and Nina [CHS20].

Previous research on T. forsythia indicates that the human-associated strains fall into two major lineages whose histories are entangled with the impact of European colonialism18,21. Lineage 1 appears to be limited to the pre-contact era Americas, as it has only been recovered from dental calculus samples from pre-contact Indigenous individuals from Mexico21, and Wichita Ancestors from Oklahoma, USA18. Lineage 2 strains have been identified in historic era (18th-19th century) individuals from Europe and South Africa, and in modern individuals from the USA, Europe, and Asia, suggesting that it is the major lineage prevalent worldwide today and likely replaced Lineage 1 after European contact18. Our phylogenetic analysis based on alignments of homozygous SNPs across the T. forsythia genome placed Tima’s [CHS31] strain in Lineage 2, although its relationship to other Lineage 2 strains could not be assessed due to limited bootstrap support (Fig. 5, Supplementary Fig. 6). The placement of Tima’s [CHS31] strain suggests that historic period African persons carried strains similar to historic populations outside of North America.

Fig. 5: Maximum Likelihood tree of Tannerella forsythia strains.
figure 5

An alignment of 26,922 homozygous, bi-allelic SNPs was used. The tree was built using IQ-TREE and the K3PU + F + ASC + R2 model. The tree was rooted using canine T. forsythia strains. Brown circles indicate branches with greater than 95% bootstrap support, estimated using 1000 ultrafast bootstraps. Strains are color-coded based on location, symbols are used to denote temporal context, and the country of origin is shown in parentheses. The strain reconstructed from Ancestor Tima [CHS31], is colored in red.

Full size image

Our phylogenetic analysis of the species P. alactolyticus using alignments of homozygous SNPs across the genome (Fig. 6, Supplementary Fig. 7) and a curated set of essential genes (Supplementary Figs. 8, 9) identified two major lineages in this species. Lineage 1 is present in the Americas and Asia and includes strains from pre-contact era Indigenous individuals from Mexico and Wichita Ancestors from the USA, historic era (17th–19th century) Japanese individuals, and modern individuals from China and the USA, with the latter being the reference strain for this species. Lineage 2 includes strains from individuals from ancient Egypt, historic era (18th–19th century) South Africa, and modern Europe, and an environmental isolate from Italy. Removing the environmental strain did not affect the tree structure, supporting its place in Lineage 2. Strains from Lima [CHS03] and Nina [CHS20] both belonged to Lineage 2, which is otherwise not found in individuals from the Americas. Within Lineage 2, Nina’s [CHS20] strain was closely related to a historic era strain from the U.K., whereas Lima’s [CHS03] strain was closely related to a historic era strain from South Africa, suggesting it was likely brought over to the Americas during the Trans-Atlantic Slave Trade.

Fig. 6: Maximum Likelihood tree of Pseudoramibacter alactolyticus strains.
figure 6

An alignment of 3,083 homozygous, bi-allelic SNPs with a maximum of 10% missing data was used. The tree was built with IQ-TREE using the K3Pu+F + ASC + R2 model. A strain reconstructed from a wild-borne chimpanzee was used as the outgroup. Brown circles indicate branches with greater than 95% bootstrap support from 1,000 ultrafast bootstrap replicates. The color of each strain corresponds to geographic origin and the symbols equate to temporal context. Strains reconstructed from Ancestors Lima [CHS03] and Nina [CHS20] are colored in red.

Full size image

Analysis of enamel strontium isotopes for Lima [CHS03] and Nina [CHS20] indicated that these Ancestors were probably not born in Africa, and thus likely represent second or later generations of African descendants whose ancestors were brought to the Americas through forced migration44. Despite their local origin, these Ancestors do not carry strains of T. forsythia and P. alactolyticus native to North America which could have been acquired through contact with Indigenous communities in the vicinity or from the local environment. In Lima’s [CHS03] case, at least one of these oral strains likely originated in Africa.

The finding of strains closely aligned with those from historic and modern African or European populations suggests the Ancestors either acquired them from African persons through vertical transmission over generations through caregivers60,61 or via frequent contact with other persons of African descent and/or enslavers or other persons of European descent. For the latter, this could possibly manifest through cultural practices or demographic processes17. The exact transmission dynamics can be difficult to disentangle in archaeological contexts, since oral strains can be acquired through both vertical and horizontal transmission over the course of an individual’s lifetime62. However, the competitive interplay between different strains may result in discordant evolutionary outcomes for different species within the same oral ecology.

Our phylogenetic analysis indicates differences in the evolutionary trajectories of two oral microbes, T. forsythia and P. alactolyticus, in the Americas. In case of T. forsythia, the non-American lineage seems to have replaced the pre-existing American lineage after contact, whereas, in the case of P. alactolyticus, the pre-existing American lineage persists today. However, wider sampling and genomic data for modern strains from the Americas will help provide a better understanding of how contact with Europeans and the subsequent forced Trans-Atlantic migrations influenced the strain dynamics of different oral microbes.

Analysis of microbial functional profiles in the context of diet

Although dietary practices can influence oral microbiome composition63,64,65,66,67,68, recent literature has highlighted the complexities of identification and authentication of dietary DNA in low-coverage ancient oral metagenome data16,69. We attempted to detect eukaryotic DNA sequences in the Ancestors’ oral metagenomes by screening the analysis-ready reads using Kraken270 and a custom database comprising prokaryotic, mitochondrial, and plastid genomes. However, we were unable to authenticate damage patterns for these eukaryotic DNA sequences, preventing direct determination of dietary constituents (Supplementary Fig. 10). Therefore, we employed an indirect approach to infer dietary components using an analysis of microbial gene functions related to carnivorous and herbivorous diets71.

We used HUMAnN347 to characterize microbial functional profiles of the Anson Street Ancestors, along with those of pre-contact era agricultural Wichita Ancestors from the USA, historic era individuals from the Radcliffe Infirmary, United Kingdom, and modern Spanish individuals to contextualize possible dietary constituents across subsistence practices. We assessed the presence of 40 enzymes previously identified71 in gut metagenomes to be differentially abundant depending on diet (Supplementary Data 11)69. Resource availability strongly drives the functional potential of a microbiome, with functional diversity often being decoupled from taxonomic diversity in metabolically niche environments72,73,74, supporting the assumption that enzymes linked to meat consumption in gut microbiomes can be identified in oral microbiomes, regardless of the originating body site.

Of these 40 enzymes, we identified 38 that could be linked to 66 KEGG Orthologous (KO) categories measuring plant or meat consumption. A total of 31 KO categories were present at greater than two counts per million (cpm) in at least 75% of dental calculus samples. KO categories measuring plant consumption did not show significant variability across the comparative populations (p-value > 0.07), although there was a distinct difference in the abundance of KO categories associated with meat consumption (p-value < 8 × 10−12) (Fig. 7; Supplementary Fig. 11). Overall, the Anson Street Ancestors had an average abundance of 40.8 cpm of meat-associated KO categories compared to 67.8 cpm (p-value < 9 × 10−6) for the Wichita, 60.4 cpm (p-value < 7 × 10−5) for the historic Radcliffe, and 105.6 cpm (p-value < 9 × 10−11) in the modern Spanish individuals. This finding implies that, in comparison to the other analyzed populations, the Ancestors did not consume large amounts of animal protein.

Fig. 7: Functional analysis of animal protein consumption.
figure 7

HUMAnN3 gene family abundances aggregated based on high-meat or high-plant consumption. Pairwise Wilcoxon rank sum tests for significance with p-values < 0.05 are indicated below the graph by asterisks. The number of genes in each population are displayed above each bar and detailed statistics are outlined in Supplementary Data 18.

Full size image

The lower consumption of animal protein by individuals of African descent in enslavement conditions is supported by historical, archaeological, and isotopic research75. Archival records indicate that diets primarily consisted of vegetables and grains, many grown by the enslaved individuals themselves, and supplemented with locally foraged wild game and fish76. Faunal analyses from archaeological sites in North America and the Caribbean also support these findings, displaying evidence of consumption of wild game and fish, and fewer prime cuts of stock animals like cows77,78,79. Nitrogen stable isotope ratios further indicated a diet lower in meat or possibly higher in wild game and fish for enslaved individuals80,81. An ongoing analysis of nitrogen isotopes for the Anson Street Ancestors will assist in evaluating the link between diet and microbial functional profiles. Additionally, analyzing samples from non-enslaved populations in South Carolina is crucial for discerning whether the observed microbial pattern is due to enslavement conditions or instead represents a regional pattern of dietary practices.

Conclusions

Employing a community–engaged approach, this study characterized the oral microbiome profiles of the Anson Street Ancestors to understand their lived histories, health, and diet. This study also represents the largest assemblage of oral microbiome data from colonial era African individuals published to date. Our analysis revealed no strong associations between oral microbiome composition, oral health, and childhood residency, suggesting the relative stability of oral community structure during a person’s lifetime. Additionally, the Ancestors’ oral community profiles were similar to those of other archaeological individuals, but differed from those of individuals from the modern era, indicating that oral communities have shifted over time. Moreover, our investigation into the phylogenies of Tannerella forsythia and Pseudoramibacter alactolyticus revealed that the lineages of these species carried by the Ancestors correspond to those found in European and African populations.

These findings offer valuable insights into oral microbial histories which, in turn, inform our larger understanding of oral health and diet for enslaved African or African descendant populations in the U.S. South. Ultimately, this study significantly contributes to our growing understanding of the oral microbial histories, shedding light on their oral health and dietary patterns, as well as emphasizes the pivotal role of community engagement in oral microbiome research.

Methods

Ethics and community engagement

The ASABG Project is a community-based effort to understand the lives of 36 Ancestors of African descent in conversation and collaboration with people living in the Charleston region43. Public presentations describing what can be learned from the analysis of dental calculus were given during community events prior to starting laboratory analyses to ensure adequate time for people to learn and provide feedback about the study. We made additional public presentations to share the progression of preliminary findings, as well the final results presented here. The contents of this manuscript were subsequently shaped following these conversations, which primarily concerned questions about oral health and diet of the Ancestors.

Oral pathological analysis

Oral pathologies and modifications were assessed for all individuals with intact dentition as outlined in Fleskes et al. 44. Specifically, features associated with oral health (carious lesions, abscesses, tooth loss, and infection), wear (generalized, occlusal, and modifications), and enamel defects (enamel hypoplasia and staining) were noted for each tooth containing dental calculus and the available dental arcade for the individual.

Dental calculus sampling

Calculus samples were collected from the teeth of 14 Ancestors in May 2019 at Brockington & Associates in Mount Pleasant, South Carolina (Supplementary Fig. 1) following guidelines from Velsko and Warinner83. Dental scrapers were decontaminated between each sample by wiping all tools with 10% bleach followed by 70% ethanol. Sterile gloves were worn and replaced between each sampling. Calculus samples were stored at room temperature until DNA extraction. The calculus samples collected ranged between 0.2–10.6 mg in weight (Supplementary Data 1). Samples from multiple teeth were combined for three Ancestors in order to have sufficient amounts for DNA extraction.

Shotgun sequencing

All laboratory work was conducted at the Laboratories of Molecular Anthropology and Microbiome Research, University of Oklahoma (Norman, USA), following established workflows for aDNA research84. The calculus samples were irradiated for 1 minute on each side using a UV Crosslinker (VWR) and washed with 1 mL of 0.5 M EDTA for 15 minutes to remove surface contaminants. A fresh solution of 1 mL 0.5 M EDTA solution and 100 μL Proteinase K (Qiagen) was added, and the calculus samples were allowed to decalcify over a 4-day period. Following calculus decalcification and digestion, a MinElute column-based purification (Qiagen) was conducted, as described in Ozga et al. 85, with DNA being eluted in 60 μL Buffer EB (Qiagen) pre-heated to 65°C. An extraction blank (PCR-grade water, GeneMate) was processed along with the calculus samples to monitor for laboratory contamination. DNA extracts were quantified using the Qubit High Sensitivity dsDNA Assay (Life Technologies).

A 30 μL aliquot of DNA extract was partially treated with uracil-DNA-glycosylase86 and built into a double-stranded DNA library as given in Carøe et al. 87. A library blank (PCR-grade water, GeneMate) was processed during library preparation to monitor for laboratory contamination. Libraries were indexed with two unique eight bp barcodes using the Kapa HiFi Uracil+ enzyme (Kapa Biosystems) and purified using the Sera-Mag SpeedBead clean-up protocol (Thermo Scientific), as described in Honap et al. 18. Library fragment size and concentration was assessed using the D1000 assay on the TapeStation (Agilent).

Library samples from two Ancestors showed negligible proportions of DNA fragments in the targeted size range of 150–500 bp and were thus excluded from further processing. The remaining 12 libraries were pooled in equimolar amounts. The resulting pool was concentrated using the Vacufuge (Eppendorf) to a volume of 30 μL and used as input for the PippinPrep (Sage Systems) to perform size-selection for a target range of 150–500 bp to remove adapter dimers. The resulting eluate was shotgun-sequenced using the Illumina HiSeq X and paired-end 2 × 150 bp sequencing at Admera Health (New Jersey, USA).

Sequence read processing

Raw reads were processed using AdapterRemoval v2.3.388 with the following options: –trimns, –trimqualities, –minquality 30, and –minlength 30. Reads with a minimum overlap of 10 bases were merged.

Human genome filtering

To assess endogenous human DNA content, filtered, merged reads were aligned to the human reference genome (hg19) using Bowtie v2.4.189, keeping both aligned and unaligned reads. Using SAMTools v1.16.190, the resulting SAM files were converted to BAM format, sorted, and split into BAM files of either aligned or unaligned reads. For the hg19-aligned BAM file, duplicate reads were removed using DeDup v0.12.891, and MapDamage v2.2.146 was used to assess DNA damage patterns. The BAM files containing reads not aligned to hg19 were converted to FASTQ files using the bam2FastQ tool from BamUtil v1.0.1592; these reads, hereafter referred to as ‘analysis-ready’ reads, were used for subsequent analysis.

Microbial taxonomic screening and authentication

Species-level taxonomic screening was conducted using the analysis-ready reads and MetaPhlAn v4.0.147 with the –minreadlength 30 option. The 25 most abundant microbial species, in terms of average relative abundance, were visualized using a heatmap generated with hclust2.py93. The origin of the detected species was cross-validated using the Human Oral Microbiome Database94, NCBI Taxonomy Browser, and other publicly available sources. Species likely to be of environmental origin were removed.

SourceTracker v2.0.148 was used to assess oral microbiome preservation and potential extraneous contamination by converting the species-level relative abundance tables from MetaPhlAn4 into counts. The Ancestors’ calculus samples were used as the ‘sink’, whereas samples of modern dental calculus14, sub- and supra-gingival plaque95, feces from traditional hunter-gatherer and urban industrialized populations64,96,97, skin98, and soil99 were processed using the same parameters and used as ‘sources’. Gibbs sampling was performed using the following parameters: source and sink rarefaction depth = 20,000, alpha1 = 0.01, and alpha2 = 1.0. The results were visualized as pie charts generated using a custom script.

Comparative datasets

To assess temporal and geographic changes in oral microbial diversity and composition, we used the Ancient Metagenome Directory100 to select publicly available oral metagenomic datasets. These ranged from ancient to modern contexts, including modern Spain14,16, historic Asia20, historic Africa16, historic United Kingdom14, ancient Africa16,101 and pre-contact North America18 (Supplementary Data 7). For this study, time period designations for modern contexts include samples collected from present day populations, historic period from individuals living 300–200 YA, post-contact around 700 YA, and ancient from 14,500–1600 YA. All datasets were processed using similar parameters as the Ancestors’ dental calculus samples.

Principal component analysis

To identify correlations between sample metadata and taxonomic composition, we carried out PCA using R v.4.3.1. The MetaPhlAn4 species level table was filtered to include only those species present at greater than 0.05% relative abundance in at least 10% of samples. We used the phyloseq102 and mix0mics packages103 and applied a centered-log ratio (CLR) transformation suitable for compositional data. PCA plots were generated using ggplot2104. Euclidean distances were calculated from the first principal component using the vegan package105. The Adonis2 function was used to perform PERMANOVA testing to generate R2, F, and P values between the generated distance matrices and sample metadata. Bi-plots were generated following Velsko et al. 15, with the top 10 species showing the highest and lowest negative loadings identified and plotted using a correction factor in ggplot2104.

Microbial genome reconstruction

Analysis-ready reads were mapped to the reference genomes of 173 microbial species identified using MetaPhlAn493 (Supplementary Data 10), using BWA aln v.0.7.15-r1140106 with the following parameters: -l 1024, -q 37, -n 0.1. The resulting SAM files were converted to the BAM format, unmapped reads were filtered out, and the BAM files were sorted using SAMTools v1.16.190. DeDup v0.12.891 was used to remove duplicate reads. MapDamage v2.2.146 was used to assess DNA damage patterns and rescale the BAM files to reduce quality scores of likely damaged positions. Qualimap v.2.2.1107 was used to generate coverage statistics.

Variant calling was conducted if the mapped reads were authenticated using MapDamage, and at least 30% of the reference genome was covered at least five-fold. Prior to variant calling, the trimBam option in BamUtil v1.0.1592 was used to remove two bases from each end of the reads to avoid false variant calls due to the presence of DNA damage. The resulting clipped BAM file was used as input for SAMTools mpileup with the -aa option to output all sites. A VCF file was generated using VarScan v2.3.9108 mpileup2cns with the following parameters: minimum coverage = 5, minimum reads to call a variant = 3, minimum average quality = 37, minimum variant frequency = 0.2, minimum frequency for homozygotes = 0.9, p-value = 1, and strand filter = 0.

Microbial phylogenetic analysis

To ascertain the phylogenetic relationships of T. forsythia and P. alactolyticus strains reconstructed from the Ancestors, publicly available genomic data for ancient dental calculus and modern oral sources (calculus and plaque) were used as comparative datasets. Additionally, complete or contig-level genomes were acquired from the National Center for Biotechnology Information (NCBI). For the latter, paired-end reads were simulated using wgsim109 with the following parameters: -1 = 150, -2 = 150, -d = 100, -r = 0, -R = 0, -X = 0, and -e = 0. For all datasets, reads were mapped to the appropriate reference genomes and variant calling was conducted in a manner similar to that for the Ancestors.

Samples were included in the phylogenetic analysis provided they fulfilled four criteria: (1) more than 30% of the genome was recovered at a minimum of five-fold coverage; (2) the ratio of heterozygous SNPs to homozygous SNPs was less than 1, signifying presence of a single, dominant strain; (3) they exhibited characteristic damage patterns for ancient DNA; and (4) genomes could be retrieved from comparative datasets. Summary statistics related to mapping and variant calling are provided in Supplementary Data 12 and 13. While we were able to reconstruct and authenticate partial or near-complete genomes of several oral species from the Ancestors’ samples, we focused our phylogenetic analysis only on Tannerella forsythia and Pseudoramibacter alactolyticus due the parameters described above.

To build the alignment, VCF files for selected samples were merged using bcftools v1.5110 with the -m all option. For SNP-based phylogenies, bcftools view was used to output only bi-allelic SNPs (options: –type snps, -m 2, -M 2). Insertion-deletions and SNPs occurring in repetitive regions, rRNAs, tRNAs, mobile and insertion elements, and phage-related genes were removed using vcftools v.0.1.17111. The genomic regions excluded for P. alactolyticus are presented in Supplementary Data 14, whereas those for T. forsythia are described in Honap et al. 18. FASTA files containing the final SNP alignments were generated from the VCF file using a publicly available script112 and used for building phylogenetic trees. Maximum Likelihood SNP trees were built using IQTREE v1.6.12113 with the following options: -m MFP, -bnni, and -alrt. Branch support was assessed by performing 5000 replicates of ultrafast bootstrap approximation114.

For P. alactolyticus, a multi-gene phylogeny was built based upon an alignment of 205 genes (Supplementary Data 15) annotated as “essential” in this reference genome according to The Bacterial and Viral Bioinformatics Resource Center115. A Maximum Likelihood gene tree was built using IQTREE and similar parameters as given before. Maximum Parsimony trees for both SNP- and gene-based alignments were built using MEGA-X116, with the SPR algorithm and 1000 bootstrap replicates. All trees were visualized using FigTree (http://tree.bio.ed.ac.uk/software/figtree/).

Functional analysis

Analysis-ready reads were screened for metabolic potential using HUMAnN v3.647. This pipeline taxonomically assessed the metagenomes using MetaPhlAn493 (database mpa_vJan21_CHOCOPhlAnSGB_202103) and characterized gene family groups through pangenome nucleotide alignment and translated alignment to the UniRef50117 database to produce taxonomically informed functional profiles. The output includes gene families reported in reads per kilobase (RPK) and pathways defined by MetaCyc118 reported as the total abundance of genes catalyzing reported reactions. Gene families were used for comparative analyses and merged with those acquired from pre-European contact Wichita samples18, historic samples from the Radcliffe Infirmary (United Kingdom) and modern samples from Spain14. Gene families reported as RPK, or relative gene copy number, in each community were normalized to copies per million (cpm) to account for any variation in sequencing depth and unmapped categories were excluded (Supplementary Data 19). Normalized gene families were regrouped to provide KEGG orthologous (KO) categories119,120,121, excluding ungrouped gene families, and unstratified tables were filtered to include only those functional categories present at a minimum of 2 cpm in at least 75% of samples.

Dietary assessment

An initial direct assessment of dietary constituents through screening for eukaryotic DNA was attempted using Kraken270 with a custom database of mitochondrial and plastid genomes from NCBI, as well as representative fungal, viral, and protozoan genomes from NCBI and bacterial and archaeal genomes from the Genome Taxonomy Database (GTDB) v214.1122,123,124,125 to ensure a diverse reference database for competitive read mapping126. The database was built and the samples assessed using default parameters (k-mer length of 35 and minimizer length of 31). Kraken2 classification was followed by Bracken v2.6 analysis, a Bayesian re-estimation tool that accompanies Kraken tools and predicts abundance at specified taxonomic levels127. The Bracken database was built with both a k-mer size and read length of 35. BWA v0.7.17106 was then used to align the analysis-ready reads to the 20 most abundant eukaryotic species detected using Bracken. Coverage, read length, and damage profiles of alignments with >500 unique reads were used to assess the validity of the presence of eukaryotic dietary constituents (Supplementary Data 16-17; Supplementary Fig. 10).

For the indirect assessment, the annotated KEGG abundance tables were screened for evidence of dietary constituents. KO categories with Enzyme Commission (E.C.) numbers enriched in herbivorous and carnivorous mammalian populations71 were extracted to infer meat consumption (Supplementary Data 11). Kruskal-Wallis tests for statistical significance and Pairwise Wilcoxon rank sum tests with false discovery rate correction128 were performed on each category in R and p-values were reported (Supplementary Data 18).