Paleogenomic insights into the red complex bacteria Tannerella forsythia in Pre-Hispanic and Colonial individuals from Mexico

Abstract

The ‘red complex’ is an aggregate of three oral bacteria (Tannerella forsythia, Porphyromonas gingivalis and Treponema denticola) responsible for severe clinical manifestation of periodontal disease. Here, we report the first direct evidence of ancient T. forsythia DNA in dentin and dental calculus samples from archaeological skeletal remains that span from the Pre-Hispanic to the Colonial period in Mexico. We recovered twelve partial ancient T. forsythia genomes and observed a distinct phylogenetic placement of samples, suggesting that the strains present in Pre-Hispanic individuals likely arrived with the first human migrations to the Americas and that new strains were introduced with the arrival of European and African populations in the sixteenth century. We also identified instances of the differential presence of genes between periods in the T. forsythia ancient genomes, with certain genes present in Pre-Hispanic individuals and absent in Colonial individuals, and vice versa. This study highlights the potential for studying ancient T. forsythia genomes to unveil past social interactions through analysis of disease transmission. Our results illustrate the long-standing relationship between this oral pathogen and its human host, while also unveiling key evidence to understand its evolutionary history in Pre-Hispanic and Colonial Mexico.

This article is part of the theme issue ‘Insights into health and disease from ancient biomolecules'.

1. Introduction

Paleopathological evidence and ethnohistorical descriptions have traditionally served as the main sources for understanding population health in the past [1]. They provide rich information about the human context of diseases and the spatio-temporal variation that affect disease dynamics [2]; however, these approaches provide limited information regarding the specific pathogens associated with past diseases. The use of biomolecular tools, such as the retrieval and analysis of ancient DNA (aDNA), now provides new sources of evidence for understanding disease environments in the past.

In particular, the recovery of high-throughput sequencing and analysis of pathogen DNA from archeological remains have revealed a number of complex evolutionary patterns of the causative agents for various infectious diseases that were not apparent when solely considering historical and archeological information [3–8]. Most research in this area has focused on ancient pathogens from European samples, with a particular emphasis on the bubonic plague bacterium Yersinia pestis [5,8–10]. Only a handful of studies have explored the genetic makeup of ancient pathogens from the Americas [4,6,7], a region of interest considering the dramatic impact that newly introduced pathogens had on Indigenous populations after European contact. This research builds on these previous studies with a diachronic study of ancient pathogens from across different regions in Central Mexico.

Teeth are often the first-choice substrate for aDNA extraction as they are known to be a good source of endogenous aDNA [11] as well as ancient pathogen DNA [12], particularly for oral pathogens [3,13,14]. Furthermore, dental calculus (i.e. mineralized biofilm) has recently been recognized as a valuable reservoir for microbial, dietary and host DNA [13,15]. This is important considering that oral infections represent some of the most common human diseases worldwide [16], have affected human populations throughout their evolutionary history [17] and increased in frequency with the advent of agriculture [18,19]. In addition, it is also known that oral infections can cause severe harm at the systemic level when left untreated [20].

Periodontitis (also known as periodontal disease) is a severe type of oral infection that can cause major damage to the alveolar bone [21,22]. The manifestation of this condition is the result of different bacterial complexes in the oral environment. The most pathogenic is the ‘red complex', which includes the Gram-negative bacteria Treponema denticola, Porphyromonas gingivalis and Tannerella forsythia [23]. These bacteria act in synergy causing the destruction of alveolar bone and soft tissues, which can ultimately lead to tooth loss if untreated [24]. Furthermore, the bacteria produce endotoxins that can cause systemic health conditions such as vascular disorders [25,26] and respiratory tract infections, among others [20]. In particular, T. forsythia is identified as the main contributor to the development of periodontitis, since it is the only member of the ‘red complex' with a distinctive glycosylated S-layer that promotes adherence to the host's gingival cell surfaces [27] and attenuation of the immune response [28,29]. The relevance of understanding the molecular principles of this disease is well acknowledged in the clinical context [30]; however, it has received less attention from an evolutionary standpoint. Although the recovery of T. forsythia DNA from ancient humans [3,31] and archaic hominins [32] suggests that this pathogen has infected humans for thousands of years, no study to our knowledge has leveraged genetic information from ancient teeth or dental calculus to explicitly explore the molecular properties of this infectious condition from a spatio-temporal framework. Therefore, a more in-depth characterization of the genome of T. forsythia at different points in time and in different geographical regions can shed light on its evolutionary history and the coevolutionary dynamics between this oral pathogen and humans.

Since the European colonization of the Americas imposed one of the most dramatic social, cultural and demographic transitions in human history, exploring the genetic makeup of oral pathogens spanning this shift is of deep interest to unveiling the evolutionary mechanisms underlying pathogenicity and adaptation to new environments. As a first approach to address this issue, we characterized partial T. forsythia genomes from twelve individuals spanning the Pre-Hispanic and Colonial periods in present-day Central Mexico and combined these data with other ancient [3,31] and modern [33] strains. While the first contact between Europeans and different Native groups in the Mexican territory varied geographically, we use the term ‘Colonial' to refer to the period between 1519 and 1810 CE, which is the standard for Mesoamerican archaeology and the sites included in this study [34].

To retrieve ancient genomic data for this oral pathogen, we used a hybridization capture-enrichment design targeting 234 genes to reconstruct partial genomes (approx. 259 Kb) of T. forsythia for seven Pre-Hispanic and five Colonial individuals where the pathogen was identified. This allowed us to assess the phylogenetic relationship between Pre-Hispanic and Colonial strains, as well as variations between these and the other ancient and modern strains.

2. Identification of Tannerella forsythia in Pre-Hispanic and colonial individuals from Mexico

We screened 53 aDNA Illumina libraries (49 from teeth and 4 from dental calculus; electronic supplementary material table S1) by sequencing at low depth in the NextSeq550 platform, producing between approximately 2 million and approximately 20 million paired-end reads per library (electronic supplementary material table S6). Twenty samples belong to Pre-Hispanic (540 BCE–1519 CE) individuals from the modern states of Guanajuato, Queretaro, Tlaxcala and Veracruz and 33 samples to Colonial individuals from Mexico City and Queretaro (figure 1). Detailed information for each sample is found in electronic supplementary material table S1. Sequenced reads that did not map to the human reference genome were taxonomically classified using Kraken2 [35] through comparison to a database composed of whole bacterial, archaeal and viral genomes from the NCBI's Reference Sequence Database (see §7). As a spatial and temporal reference, we also included in our analysis data from four ancient (950 CE–1850 CE) European individuals [3,31] previously reported as positive for ‘red complex' bacteria.

Figure 1.

Geographic locations and temporal periods of ancient individuals included in this study. (a) Study sites with ancient individuals positive for T. forsythia. (b) Map of Mexico indicating the position of the archeological sites included in this study. The time periods of the samples analysed are indicated by colour. Yellow: Pre-Hispanic, Turquoise: Colonial. The number of samples that are positive for T. forsythia are indicated in brackets for each archeological site location.

The per cent of reads assigned to a taxon by Kraken2 ranged from 1% to 34%, with a predominant component of environmental and human microbiome bacteria in varying proportions (figure 2 and electronic supplementary material, table S2). We identified the presence of ‘red complex' bacteria; T. forsythia (1065 to 307 840 reads); P. gingivalis (14 to 7422 reads) and T. denticola (140 to 6807 reads) in twelve individuals: seven Pre-Hispanic and five Colonial (figure 1, tables 1 and 2, electronic supplementary material, table S2). Notably, T. forsythia was the dominant bacterium in this analysis in comparison to the other two members of the ‘red complex'. By contrast, the taxonomic profiling of reads from twelve DNA extraction and library preparation blanks (negative controls) assigned a low number of reads (2 and 6) to T. forsythia 92A2 in two blanks (electronic supplementary material table S4; see also electronic supplementary material, Discussion), giving us confidence that our results were unlikely to be false positives. Moreover, when only considering teeth, only 8 out of the 49 libraries had hits to T. forsythia, while all four dental calculus samples did. When contrasting the frequency of T. forsythia-positive samples per period, we observe 7 out of 20 samples for the Pre-Hispanic period and 5 out of 33 samples for the Colonial period.

Figure 2.

Red complex bacteria in Pre-Hispanic and Colonial individuals. Relative abundance as determined by Kraken2 [35] of the main taxa observed in each ancient individual in this study. ‘Red complex’ bacteria are coloured in shades of red, reads assigned to T. forsythia are distinguished by the most intense red colour. D = dentin sample and DC = dental calculus sample. ‘Ancient European’ individuals correspond to previously published data; G12 and B61 [3], CS21 and CS40 [31]. The main taxa categorization was based on references found in table S3.

Table 1.

Characteristics of the ancient individuals analysed in this study.

period	sample ID	substrate	date	archeological site	reference
Pre-Hispanic	CA-13	dentin	770-400 BCE*	Cañada de La virgen, Guanajuato, Mexico	this study
Pre-Hispanic	TO-2417Q	dentin	1257 ± 30 CE*	Toluquilla, Queretaro, Mexico	this study
Pre-Hispanic	TO-2417 J	dentin	900 BCE	Toluquilla, Queretaro, Mexico	this study
Pre-Hispanic	TO-333O	dental calculus	700 ± 50 BCE*	Toluquilla, Queretaro, Mexico	this study
Pre-Hispanic	VE-42	dental calculus	900 BCE–1200 CE	Tabuco, Veracruz, Mexico	this study
Pre-Hispanic	TLA-01	dental calculus	1451–1523 CE* †	Tepeticpac, Tlaxcala, Mexico	this study
Pre-Hispanic	TLA-22	dental calculus	1431–1479 CE* †	Tepeticpac, Tlaxcala, Mexico	this study
Colonial	CO-09	dentin	1700–1900 CE	Temple of the Immaculate Conception ‘La Conchita’, Mexico City, Mexico	this study
Colonial	CO-20	dentin	1700–1900 CE	Temple of the Immaculate Conception ‘La Conchita’, Mexico City, Mexico	this study
Colonial	HSJN-194	dentin	1472–1625 CE*	Hospital San Jose de los Naturales, Mexico City, Mexico	this study
Colonial	HSJN-240	dentin	1442–1608 CE*	Hospital San Jose de los Naturales, Mexico City, Mexico	this study
Colonial	UAQG	dentin	ND	Ex College of San Ignacio de Loyola, Queretaro, Mexico	this study
Ancient European	G12	dental calculus	950–1200 CE*	Dalheim, Germany	Warinner et al. [3]
Ancient European	B61	dental calculus	950–1200 CE*	Dalheim, Germany	Warinner et al. [3]
Ancient European	CS21	dental calculus	1770–1855 CE*	Radcliffe Infirmary Burial Ground collection, UK	Velsko et al. [31]
Ancient European	CS40	dental calculus	1770–1855 CE*	Radcliffe Infirmary Burial Ground collection, UK	Velsko et al. [31]

^†López Aurelio and Santacruz Cano Ramón, Anexo 2: Fechamientos, Proyecto Arqueológico Tepeticpac, 2020. Informe de Actividades de Campo y Análisis de Materiales 2012–2018, Documento en Archivo Técnico del INAH, Ciudad de México.

C14 dating is indicated by asterisks (*).

ND, Not determined.

Table 2.

Identification of “red complex” bacteria in ancient individuals from Mexico.

			Tannerella forsythia		Porphyromonas gingivalis		Treponema denticola
period	sample	total reads*	Kraken (count reads)	uniquely mapped reads	Kraken (count reads)	uniquely mapped reads	Kraken (count reads)	uniquely mapped reads
Pre-Hispanic	CA-13	4 983 380	2 303	1 733	422	236	890	464
Pre-Hispanic	TO-2417Q	7 070 465	14 559	18 867	4 490	7 422	2 760	2 767
Pre-Hispanic	TO-2417J	598 299 054	2 043 487	307 840	64 218	4 786	870 528	6 807
Pre-Hispanic	TO-333O	2 378 046	32 040	11 031	12 056	4 470	3 309	1 181
Pre-Hispanic	VE-42	2 377 859	5 295	1 206	1 096	532	2 623	743
Pre-Hispanic	TLA-01	2 383 724	10 998	3 268	2 445	737	3 235	1 069
Pre-Hispanic	TLA-22	2 383 724	7 464	1 992	2 157	263	2 597	629
Colonial	CO-09	19 446 371	40 547	6 796	8 852	1 779	19 777	3 920
Colonial	CO-20	15 681 711	5 493	4 344	1 323	920	876	767
Colonial	HSJN-194	5 321 515	8 650	1 065	736	4 600	2 641	1 803
Colonial	HSJN-240	4 773 622	1 770	4 600	616	1 065	2 136	1 305
Colonial	UAQG	7 329 193	3 275	2 438	31	14	305	140
Ancient European	G12	47 043 719	407 359	268 432	65 957	43 065	127 287	6 7028
Ancient European	B61	46 633 826	234 385	178 848	11 696	6 997	28 427	18 957
Ancient European	CS21	5 852 657	296 749	280 244	26 280	14 530	53 558	14 364
Ancient European	CS40	4 826 624	596 602	342 820	39 297	851	40 027	2 493

*Total unmapped reads to human.

Interestingly, the taxonomic profiles of the ‘red complex'-positive libraries displayed by non-metric multidimensional scaling (NMDS) reveal a clustering based on the substrate analysed (electronic supplementary material figure S1). The exceptions to this clustering pattern are samples TO-2417J and CO-09, which are dentine samples that cluster with calculus samples (electronic supplementary material figure S1, electronic supplementary material table S5). The taxa that seem to be driving this are Actinomyces sp. oral taxon 894, Olsenella and sp. oral taxon 807, Anaerolineaceae bacterium oral taxon 439, Streptococcus gordonii and T. forsythia.

We mapped the low-depth sequencing data that did not map to the human reference genome of the twelve individuals with reads assigned T. forsythia by Kraken2 to the available T. forsythia reference genome (NC_016610.1) to confirm the presence of authentic aDNA from these bacteria and to rule out potential modern contamination (see §7). This yielded between 1065 and 769 591 uniquely mapped reads (electronic supplementary material, table S6). By contrast, only three out of twelve experimental negative controls yielded between 1 and 30 reads mapping to T. forsythia (electronic supplementary material, table S6). We observed the characteristic damage and fragmentation patterns expected for aDNA libraries [36], namely the elevated C to T transitions at the molecule ends (electronic supplementary material, figure S2) and short-read lengths between 69 and 94 base pairs (bp). Additionally, the mapped reads were distributed across the entire genome and not concentrated in conserved or specific regions. Together, these observations support that the reads represent authentic aDNA from T. forsythia.

3. Capture design and targeted enrichment of aDNA libraries

To increase the genomic coverage of T. forsythia, we designed a custom set of in-solution capture-enrichment baits targeting 234 genes (258 696 bp) found in the T. forsythia RefSeq genome assembly (NC_016 610.1). The 234 genes were selected based on functional annotations in the Pathosystems Resource Integration Center (PATRIC) database [37]. These include 207 genes annotated as ‘essential', 24 genes with an annotation related to antibiotic resistance, two genes annotated as transporters and one gene annotated as a virulence factor (electronic supplementary material, figure S3). A description of the regions used for the probe design is available in electronic supplementary material, table S7.

T. forsythia DNA was enriched in 11 out of the 12 libraries and sequenced between approximately 461 thousand and approximately 58 million reads for each with two sequencing rounds on the Illumina NextSeq550. The library for individual TO-2417J was not included in the capture since it was sequenced at high depth, together with the pre-capture library for individual TO-2417Q on a NovaSeq instrument in order to reach whole-genome coverage for both samples (see §7). Upon merging the sequence data available for each sample, between 12 326 and 179 824 unique reads mapping to the reference genome were retrieved, which resulted in an average on-target depth between approximately 1.50 X and approximately 42 X . The captured libraries show between 16- and 213-fold enrichment of on-target sequences compared to the pre-capture libraries (figure 3 and electronic supplementary material, figure S4 and table S6 and S8). For individuals TO-2417Q and TO-2417J, for which we sequenced pre-capture libraries at a higher depth, we obtained whole genomes with depths of approximately 28.6 X and approximately 9.8 X, respectively. The depth for sample TO-2417Q was calculated from the merged data for all pre-capture and post-capture reads.

Figure 3.

Depth of coverage of the T. forsythia genome in three Pre-Hispanic and three Colonial samples. The depth is shown in 100 bp windows and it is considerably higher at the 234 genes that were included in the targeted capture design (red lines in outer ring). Each grey line represents approx. 2 x. From inner ring: Pre-Hispanic individuals TO-2417J, TO-333O, TLA-01; Colonial individuals: CO-20, HSJN-240, UAQG. The outmost ring represents the location of the probes. All the plots for Pre-Hispanic and Colonial individuals are shown in figure S4.

We did not find a correlation between the type of substrate (dentin or calculus) or the pre-capture endogenous content and the performance of the capture strategy; however, the pre-capture library with the shortest fragments (HSJN-194, an average of 69 bp) had the highest fold enrichment (approx. 213 fold) post-capture. Overall, there was a general trend in which pre-capture libraries with shorter average fragment lengths yielded higher fold enrichment values (electronic supplementary material table S8). However, we did not observe any significant correlation between fragment length and fold enrichment (Pearson's r = 0.10, p-value = 0.89) when considering only the four dental calculus samples, either when considering only the dentin samples (Pearson's r = −0.67, p-value = 0.09) or all samples together (Pearson's r = −0.56, p-value = 0.06).

4. Phylogenetic analysis of ancient Tannerella forsythia genomes

We selected the samples with depths above 2.5 X , which excluded Pre-Hispanic sample CA-13, to investigate their phylogenetic relationship to available T. forsythia genome-wide data. To this end, the data for the remaining 11 Pre-Hispanic (n = 6) and Colonial (n = 5) samples were merged with published ancient [3,31] and modern [38] sequence data for this oral pathogen. The comparative data included two ancient individuals (950–1200 CE) from Dalheim, Germany [3] and two ancient individuals (1770–1855 CE) from the Radcliffe Infirmary Burial Ground collection in Oxford, UK [31] (figure 1, electronic supplementary material, table S1). We also included data for two present-day T. forsythia genomes, one retrieved from a periodontitis patient [33] and T. forsythia's reference genome sequence (NC_016610.1), which belongs to strain 92A2; both modern genomes were obtained from US patients. In sum, the phylogenetic analysis considered a total of seventeen T. forsythia partial genomes, 15 ancient and 2 modern.

The raw read data from the previous studies were mapped to the T. forsythia genome with the same parameters used for our data. From the mapped data, we generated consensus sequences for 170 genes (electronic supplementary material, table S7) that were in the capture design and were also orthologous between the T. forsythia and the P. gingivalis reference genome (NZ_CP025932.1), which was used as an outgroup in the phylogenetic analysis (see §7).

The maximum-likelihood tree integrates the consensus sequences from all eighteen samples: six Pre-Hispanic, five Colonial, four ancient European, two modern T. forsythia [33,39] and P. gingivalis. Out of the 15 bootstrap support values for the bifurcations in the tree, 11 were above 75, and the lowest value was 55, consistent with an overall well-supported topology. There is a phylogenetic positioning that corresponds with the dates and geographic origin of the samples, with one clade grouping Colonial, European and modern individuals, and a second clade that includes all of the Pre-Hispanic T. forsythia samples (figure 4). In the T. forsythia Pre-Hispanic cluster, there is no clear phylogenetic proximity according to geographical location, except two samples from Toluquilla, Queretaro (TO-2417Q and TO-333O). Interestingly, TO-2417J, which is from the same site, is basal to the remaining Pre-Hispanic samples. The Colonial samples cluster together with the ancient European individuals and the two modern T. forsythia samples from the USA forming a monophyletic group. Noticeably, the reference T. forsythia genome is closest to one of the historical samples from Germany (CS40), while the second sample from this site (CS21) is basal to a group consisting of the modern and Colonial samples. The placement of the remaining samples does not seem to follow a clear temporal or geographical trend; however, individual HSJN-194 is basal to the Modern/European/Colonial clade and is closely related to individual C0–09. This is noteworthy since morphological analyses of HSJN-194 [40] and mitochondrial ancestry of both hosts (electronic supplementary material, table S1) suggests these individuals are of African ancestry. The next most basal sample from this clade corresponds to individual B61, which is one of the two most ancient (900–1200 CE) samples in the dataset and may explain its basal placement in the tree.

Figure 4.

Phylogenetic tree of 15 ancient and two modern T. forsythia genomes. A maximum-likelihood three built with 1000 bootstraps is shown. Bootstrap values are shown for each bifurcation and branch lengths are shown on top of each branch. Colours in the branch reflect the period according to the legend. Estimated or C14 dates are shown in parentheses next to the sample's names.

5. Differential presence of genes

To explore the genomic characteristics of the Pre-Hispanic and Colonial T. forsythia, we evaluated the presence or absence of the 234 genes included in our custom bait design across samples. The depth distribution across genes for each sample was contrasted (electronic supplementary material figures S5–S8), though samples CA-13 and CS21 were excluded from this analysis given the low on-target depth of coverage (1.53 X and 2.5 X , respectively). For the remaining samples, the depth values were normalized (electronic supplementary material figure S6 and electronic supplementary material table S9) and the genes with normalized values of 0 were considered as absent (See §7).

Out of the 207 genes annotated as ‘essential' in the PATRIC database, there is a differential presence of only two genes between Pre-Hispanic and Colonial individuals (figure 5, electronic supplementary material figure S6). These include the absence of pncB (BFO-2125), a nicotinate phosphoribosyltransferase in all the Pre-Hispanic individuals, except TO-333O (figure 5), while it is present in all Colonial, ancient European and modern sequences. The second gene, rpll (BFO-3362), is absent in all of the Colonial individuals except CO-09, while it is present in all Pre-Hispanic, ancient Europeans and modern sequences.

Figure 5.

Presence/absence of selected essential, virulence and antibiotic resistance genes in ancient and modern T. forsythia genomes. Normalized average depth over the 234 captured genes is shown in a heatmap for a subset of 20 antibiotic resistance genes, 1 virulence factor, 4 essential and antibiotic resistance genes, 2 essential and 2 transporter genes (rows) in 6 Pre-Hispanic, 5 Colonial, 4 ancient European and one Modern T. forsythia sequences (columns). Normalized depth of the 234 genes of all individuals analysed is shown in figure S6 and table S9.

In addition to these two genes, all of the ancient and modern T. forsythias analysed in this study lack tetQ (BFO-1235), a gene associated with antibiotic resistance by protecting the bacterial ribosome from binding tetracycline (figure 5). This pattern had been previously reported for the T. forsythia ancient genomes reconstructed from the two ancient individuals from Dalheim, Germany (G12 and B61) [3]. This gene is, however, annotated in the T. forsythia reference genome (NC_016610.1).

6. Discussion

We identified ‘red complex' bacteria (Tannerella forsythia, Porphyromonas gingivalis and Treponema denticola) in seven Pre-Hispanic and five Colonial individuals from Mexico. Although previous studies have reported the presence of these pathogenic bacteria in ancient European individuals [3,31,32,41], to our knowledge this study is the first direct account of T. forsythia in Pre-Hispanic and Colonial individuals from the Americas.

Notably, T. forsythia was the dominant bacteria in comparison to the other two members of the ‘red complex'. This bacterium is present in all four dental calculus samples tested in this study, as opposed to only 8 out of 49 ancient teeth (dentin) samples. This is consistent with previous reports of dental plaque being better at preserving aDNA from oral microorganisms than dentin [35]. Interestingly, we observed a clustering based on the substrates of ‘red complex'-positive samples in the NMDS (electronic supplementary material figure S1). Dental calculus samples showed a more condensed cluster compared to dentin samples; this pattern has been identified previously in ancient remains [41], suggesting a low influence of exogenous microbes and a more uniform microbial composition. Interestingly, we identified that two dentin samples (TO-2417J and CO-09) clustered in dental calculus samples, mainly driven by the presence of Actinomyces sp. oral taxon 894, Olsenella and sp. oral taxon 807, Anaerolineaceae bacterium oral taxon 439, Streptococcus gordonii and T. forsythia (electronic supplementary material table S5). The latter two pathogens act as ‘late colonizers', commonly found in dental calculus [31].

This observation is congruent with the ecological plaque hypothesis [42], which proposes that while certain oral pathogens, like T. forsythia, could be found in dental calculus as an innocuous member of the microbiome, their presence in dentin could be reflective of an infectious process involving periodontal disease. Further testing involving larger sample sizes with associated paleopathological indicators of periodontal disease is needed to unequivocally assess this hypothesis.

When only considering dentin, 3 out of 16 (18.7%) Pre-Hispanic individuals are positive for T. forsythia; a slightly lower proportion is observed for Colonial individuals (5 out of 33, 14.7%). Two out of the three Pre-Hispanic teeth that are positive for T. forsythia (TO-2417Q and CA-13) belong to individuals who were likely of high-ranking status according to their archaeological context. The burial of individual TO-2417Q at the site of the Toluquilla (in Queretaro State) displayed an hematite mirror and rich burial offerings (Elizabeth Mejía 2017, personal communication), while individual CA-13 was found with snail shell beads, a conch shell pectoral and pottery in the burial [43]. Both burial contexts suggest an upper social rank for these individuals (electronic supplementary material). Thus, high-ranking status could be linked to differential access to certain types of food that increased the presence of T. forsythia in these individuals [44]. It has been suggested that the excessive intake of fermentable carbohydrates results in dental biofilms experiencing extended periods of low pH [42,45], which can favour the proliferation of obligatory anaerobic species, like T. forsythia. Unfortunately, dental calculus, from which we could in principle characterize diet, was not available for these two individuals, therefore we cannot directly test this scenario. Nonetheless, these observations could inform future studies to directly test the potential associations of social status and diet with the presence of this pathogen.

To explore in depth the evolutionary relations between the different T. forsythia genomes, the depth of coverage of the ancient bacterial genomes was increased using a custom set of baits for targeted capture enrichment. This strategy, which targeted 234 genes, increased the on-target depth of coverage by one to two orders of magnitude and was effective regardless of the substrate. There is a trend of pre-capture libraries with shorter fragments (69–75 bp) that generally resulted in better fold enrichments. This is noteworthy given that the synthesized baits were 60 bp in length. Previous reports show that the length of baits influences the efficiency of enrichment by biasing the distribution of the read fragments obtained post-capture [46]. Although it remains to be tested how different factors, including bait length, can be adjusted to improve the efficiency of targeted capture-enrichment experiments, our approach proved quite useful for enriching libraries from different sources and with varying parameters of pre-capture endogenous content and average fragment lengths. Therefore, this study is a valuable precedent for future research examining oral health and periodontal disease in ancient human populations.

Characterizing a subset of genes at higher coverage allowed us to carry out phylogenetic analyses that revealed the existence of two clusters that distinguish the Pre-Hispanic and Colonial T. forsythia sequences. The Colonial samples cluster together with ancient European individuals and the two modern T. forsythia sequences (including the reference genome) from the USA, while all Pre-Hispanic individuals formed a separate monophyletic clade. This phylogenetic distribution of the sequences by period leads to interesting questions regarding the transmission dynamics of the newly introduced strain and whether a replacement of T. forsythia ‘Pre-Hispanic strains' by ‘Colonial strains' may have taken place. While no individual from the Colonial period carries a ‘Pre-Hispanic strain', our sample size is too small to unambiguously assert this and future studies with larger sample sizes are needed to explicitly test this hypothesis. It is unclear how T. forsythia is transmitted; however, other periodontal pathogens (e.g. Porphyromonas gingivalis and Actinobacillus actinomycetemcomitans) follow a cohabitation-based transmission pattern [47]. These observations allow us to formulate hypotheses about the use of this pathogen as a proxy to infer past human interactions and should be addressed by future research. Additional studies characterizing ancient and present-day T. forsythia strains from several geographical regions and time periods could draw a more complete picture of these transmission dynamics and elucidate how integrating knowledge of host ancestry and the putative geographical origins of their pathogenic strains could be used to infer human interaction networks.

When evaluating the presence/absence of the 234 captured genes, we observed a differential presence of two genes (pncB and rpll) annotated as having essential functions between Pre-Hispanic, Colonial and ancient European individuals.

The gene pncB is absent in all Pre-Hispanic genomes, except one (TO-333O). This gene is involved in NAD biosynthesis and in most bacterial species follows two pathways, one by genes nadB-nadA-nadC and a second one regulated by genes pncA-pncB [48]. While most bacteria have both pathways, some are strictly dependent on one or the other [48]. Our capture design included the three genes nadB, nadA and nadC, and we did not see absence of any of these. Therefore, we can speculate that nadB-nadC genes could be compensating for the lack of gene pncB. An additional plausible explanation is that in spite of pncB gene being annotated as essential in T. forsythia 92A2 reference genome, its enzymatic function could be compensated by other enzymes, a phenomenon that has been observed frequently among bacteria [49]. Lastly, gene rpll (ribosomal protein L9) is absent in all Colonial individuals except one (UAQG). Despite being a ribosomal protein and annotated as essential, this gene has also been found to be absent in other bacterial genomes [50].

The tetQ gene (BFO-1235) involved in resistance to antibiotic tetracycline is absent in all of the ancient and one of the modern genomes in this analysis. This finding replicates what was observed in the study of the two ancient samples from Dalheim, Germany that we included in our analyses [3]. This gene is present in T. forsythia's reference genome and has been found at varying frequencies (8–80%) among periodontitis patients in Europe [51,52] and the USA [53], while in Latin American populations, it is present at high frequencies [54]. Our observations suggest that this gene might have arisen under recent selective pressures; however, this is an open question that remains to be evaluated once a larger dataset becomes available.

To evaluate the possibility that our capture design may have been unsuccessful at targeting and enriching some genes and mistakenly marking these as absent, we leveraged the whole-genome data for T. forsythia (13 X average depth from the higher depth sequenced pre-capture library only) for individual TO-2417Q. When applying the same presence/absence criteria on the whole-genome data and contrasting the results with those from the post-capture library, the same genes are identified as absent (electronic supplementary material, figure S6). Therefore, we can rule out any considerable bias introduced by the capture efficiency that would mistakenly render specific genes as missing. In addition, any systematic error in the probes would be revealed in having the same gene as missing in all the samples from this study, which is not observed.

Overall, while limited to only the genes considered in the capture design, the patterns identified with the differential presence of genes point to some of the evolutionary changes experienced by T. forsythia during a wide temporal range. Since this pathogen has accompanied humans for at least tens of thousands of years, it is not surprising that the genome of T. forsythia co-adapted to the changing environments (diet, lifestyle, oral hygiene, etc.) experienced by their human hosts as they migrated, interacted with other populations and modified their environments. It is also notable that in this study the antibiotic resistance gene tetQ is absent in all T. forsythia genomes, except in the reference one, which could reflect an early acquisition of this gene. This also reveals how the selective pressures posed by antibiotics can differ geographically and temporarily, and that the presence or absence of antibiotic resistance genes in samples with spatio-temporal diversity can reflect the mode and tempo of such pressures [3].

In conclusion, this study provides the first account of evolutionary patterns in the ‘red complex' bacterium T. forsythia by leveraging ancient and present-day genomic data. The identification of this pathogen in Pre-Hispanic individuals from Mexico is indicative of the long-standing relationship between T. forsythia and its human host. Moreover, we were able to recover informative amounts of aDNA from this pathogen by means of a capture-enrichment strategy that considerably increased the depth of coverage for 234 informative genes. In turn, these data allowed us to dissect relevant evolutionary information and uncover the phylogenetic relationships between ancient sequences spanning the transition from Pre-Hispanic to Colonial times. This study, therefore, illustrates the potential use of ancient oral pathogen genomes as a proxy to infer patterns of social networks in the past.

Though limited by small sample size and the targeting of only a fraction of the genes in the T. forsythia's genome, our results unveil many future directions for the study of this pathogen. Further characterization of whole ancient T. forsythia genomes from different spatio-temporal contexts and diverse populations can provide valuable data to validate the usability of this pathogen as a proxy for inferring past social interactions and to characterize in depth host–pathogen coevolutionary dynamics through extended periods of time.

7. Material and methods

(a). Sample provenance

Teeth and dental calculus samples (N = 53) were obtained from the archaeological collections of four Pre-Hispanic sites and three Colonial sites. The Pre-Hispanic individuals are from Cañada de la Vírgen, Guanajuato (n = 11 teeth); Toluquilla and Ranas, Querérato (n = 1 dental calculus, n = 5 teeth); Tepeticpac, Tlaxcala (n = 2 dental calculus); and Veracruz (n = 1, dental calculus). The Colonial individuals are from the Temple of the Immaculate Conception 'La Conchita', Mexico City, Mexico (n = 7 teeth), the Hospital San Jose de los Naturales, Mexico City (n = 21 teeth) and Ex College of San Ignacio de Loyola, Queretaro (n = 5 teeth) (figure 1 and table 1). Sample HSJN-194 was radiocarbon (C14) dated at the Accelerator Mass Spectrometry Laboratory (LEMA), at the Institute of Physics, National Autonomous University of Mexico (IF-UNAM) in Mexico City. Permits to carry out DNA analyses with these samples were granted by the Archaeology Council of the Instituto de Antropología e Historia (INAH) of Mexico. Additional information on the context of the samples is available in the electronic supplementary material.

(b). DNA extraction and library preparation

Samples were processed in a dedicated aDNA laboratory at the International Laboratory for Human Genome research (LIIGH-UNAM, Queretaro, Mexico), following strict procedures to avoid modern DNA contamination [55,56]. Dental plaque samples were collected using a dental scaler. Teeth and dental calculus samples were UV irradiated (256 nm) for 3 min using a UVP CL-1000 cross-linker. Teeth were cleaned with a 1% sodium hypochlorite solution. A Dremel tool was used to remove the outer surfaces of teeth and to slice transversely at the cemento-enamel junction [57]. The roots were then covered in aluminium foil and pulverized using a hammer, as in [58]. The ancient DNA extraction for both teeth and dental calculus samples was carried out using a silica-based method [59,60], as described in [61], using approximately 200 mg of pulverized tooth or approximately 20 mg of calculus [3,41].

In order to detect contaminants in reagents or by human manipulation, extraction, library and PCR indexing blanks were processed in parallel. DNA extracts and extraction blanks were quantified with the Qubit 2.0 High Sensitivity assay [62]. For all the samples, we created barcoded (6 bp), double-stranded Illumina libraries, as previously reported [63]. Pre-capture and captured sample libraries, and 12 negative controls, carried along with DNA extraction, library preparation and PCR indexing were sequenced on an Illumina NextSeq550 (2 × 75 cycles) at the LANGEBIO's genomics core facility (National Laboratory of Genomics for Biodiversity, Irapuato, Guanajuato). Additionally, two pre-capture libraries (TO-2417Q and TO-2417J) were sequenced at higher depth to obtain whole genomes for T. forsythia on the NovaSeq instrument (S1, 2 × 100 cycles) available at the LANGEBIO's genomics core facility.

(c). Sequence data quality control and taxonomic classification

AdapterRemoval v. 2 [64] was used to process the fastq files to clip adapter sequences and merge sequence pairs (with at least 11 bp overlap). Only reads 30 bp and longer and with a quality above 33 were retained for downstream analyses. For pre-capture libraries, reads were first mapped to the human reference genome (version GRCh37, hg19) using ‘bwa aln’ option with (flags −l500 −t 16 −q 25) and the unmapped reads were then retrieved for downstream analyses. Unmapped reads from dentin, dental calculus and negative control libraries were then taxonomically binned with Kraken 2 [35], using only-classified-output and NCBI RefSeq bacterial, archaea and viral genomes as the reference database. To extract the counts assigned at the species level we used Kraken-biom (https://github.com/smdabdoub/kraken-biom). The species that were present at less than 0.02% relative abundance were removed from downstream analyses, as in [29]. Metagenomic profiles were analysed in-depth with Pavian [65] to identify known pathogens. The visualization of the relative abundance of taxonomic groups and non-metric multidimensional scaling (NMDS) was performed with the phyloseq library [66,67]. The taxonomic assignments were evaluated manually to identify known pathogens.

(d). Mapping and authenticity of ancient pathogen DNA

The reads from libraries that were T. forsythia positive were mapped to the Tannerella forsythia genome 92A2 (assembly NC_016610.1), Porphyromonas gingivalis ATCC 33277 (assembly NC_010729.1) and Treponema denticola ATCC 35405 (assembly NC_002967.9) using bwa aln option [68] (with the flags −l 1024 −n 0.03, −q 37). Clonal duplicates were removed using samtools rmdup [69]. Damage and fragmentation patterns were assessed using mapDamage2 [70] with default parameters. The coverage distribution of reads across the reference genome was visualized using Artemis [71].

(e). Capture-enrichment of Tannerella forsythia genes

A custom set of in-solution capture-enrichment baits was designed targeting 234 genes of T. forsythia based on genome assembly NC_016610.1. The 234 genes (258 696 total bp) were selected based on annotations in the Pathosystems Resource Integration Center (PATRIC) database [37] (electronic supplementary material table S5). The probes were designed by Arbor Biosciences (myBaits Custom kit, Ann Arbor, MI, USA) using their proprietary pipeline, which yielded ca. 20 000 probes, 60 bp in length, and with 6 x tiling density for each genomic region. Each bait candidate was aligned using BLAST [72] against the reference genomes of Treponema denticola (NC_002967.9), Porphyromonas gingivalis (NC_010729.1) and Tannerella forsythia (NC_016610.1). The hybridization melting temperature (Tm) was estimated for each hit assuming standard myBaits Custom kit buffers and conditions.

DNA capture was performed on the indexed libraries following the manufacturer's protocol (myBaits Kit protocol v4). Libraries were amplified with Phusion U Hot Start DNA Polymerase (Thermo Fisher Scientific), purified with SPRISelect Magnetic Beads (Beckman Coulter) and quantified with a Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA, USA). Captured libraries were amplified for 12 cycles. Amplified libraries were pooled in equimolarity and sequenced on the NextSeq550 (2 × 75 cycles). Resulting fastq files were processed and mapped as described above for the pre-capture libraries, except that for post-capture libraries we mapped directly to T. forsythia's reference genome and skipped the human genome mapping step.

We used BEDTools intersect [73] to determine the number of reads on target. The input to BEDTools intersect were the bam file and a file in bed format with the coordinates of the targeted genes; the output was a bam file with the intersecting reads. The output bam files were then processed with SAMtools ‘depth’ [69] function to calculate the genome depth and coverage of the alignments to the T. forsythia reference genome. To evaluate the success of the capture-enrichment strategy, the total numbers of unique reads mapping to the target regions before and after capture were compared (electronic supplementary material table S6). Since different numbers of reads were obtained in the pre- and post-capture sequencing runs, an equal number of reads (the lowest between the pre- and post-capture runs) for each individual were subsampled using seqtk (https://github.com/lh3/seqtk). The genome coverage was calculated in 100 bp windows and was plotted using Circos [74].

(f). Phylogenetic analysis

Consensus sequences for six Pre-Hispanic Mexico (CA-13, TO-2417Q, TO-2417J, TO-333O, VE-42, TLA-01, TLA-22); five Colonial Mexico (CO-09, CO-20, HSJN-194, HSJN-240, UAQG); four ancient European (G12, B61, CS21, CS40) and two publicly available modern T. forsythia genomes (modern and reference) [33,39] were generated using ANGSD [75] (flags -doCounts 1 -doFasta 2), for 170 orthologous genes between T. forsythia and P. gingivalis ATCC 33277 (NC_010729.1). This set of 170 genes was defined by selecting the orthologous genes between both species in Orthologs Database (OrtholugeDB) [76] that were also present in the custom capture-enrichment design. Gene consensus sequences were concatenated and aligned using MAFFT [77], with 1000 as the maximum number of iterations. The poorly aligned sequences were trimmed using trimAl [78], with a strict set of parameters, based on an automatically selected similarity threshold. trimAl uses the residue similarity scores distribution from the MSA (multiple sequence alignments) and selects the values at percentiles 20 and 80 of the alignment length. The lower and upper boundaries allow the retention of the 20% most conserved columns in the alignment, whereas the 20% most dissimilar columns are discarded [78]. RaXML [79] was used to construct a maximum-likelihood phylogeny based on the multiple sequence alignment, with the GTRGAMMA substitution model [79] and 1000 bootstrap replicates.

(g). Gene presence/absence

Reads were re-mapped to the T. forsythia genome (assembly NC_016610.1) using bwa aln [68] with the flags −l 1024 −n 0.03, −q 0 as in [6]. The rationale is when BWA identifies a read that maps to more than one location it randomly selects one and assigns it a 0 mapping quality, so when filtering by quality one it would disregard these reads, which in turn could increase the chances of a gene with paralogous sequences in the genome being considered as absent. Coverage at the 234 genes selected for enrichment was calculated using the BEDTools [73]. We used a min–max normalization to reduce the scale values to 0 and 1 and make data comparable between individuals. The following formula was used:

$$Yi=\hspace{0.17em}\frac{Xi\hspace{0.17em}-\hspace{0.17em}\text{min}\hspace{0.17em}(x)}{\text{max}(x)\hspace{0.17em}-\hspace{0.17em}\text{min}(x)},$$

where min and max are the minimum and maximum values in X given its range. So Xi converts to Yi. We considered genes to be absent only if they had a normalized depth value of 0. These criteria are equivalent to defining a gene as absent if its depth is zero; however, the normalization step allows us to contrast samples with very different average depth ranges across genes (figure 5, figures S5–S8). The normalized depth of each gene in the ancient and modern genomes in our dataset was plotted using the pheatmap library [80] in R [67].

Ethics

The handling of the archeological remains analysed in this study was based on the Protocols for the Conservation and Protection of Cultural Heritage of Mexico. The permissions for the analysis of samples were acquired in accordance with the regulations determined by the Archaeological Council of the National Institute of Anthropology and History through the Ministry of Culture in Mexico. The official notice number of the permissions granted to analyse the individuals from the Hospital San Jose de los Naturales, Temple of Immaculate Conception 'La Conchita' and Tepeticpac, Tlaxcala collections are 401.1S.3-2018/1373, 401.1S.3.2020/1310 and 401.1S.3/2017/1995, respectively. The permit to analyze the individual from Tabuco, Veracruz was included in the project IN302219 PAPIIT-DGAPA-UNAM, headed by Carlos Serrano Sanchez.

Data accessibility

Raw data (fastq files) are available upon request. Mapped reads (bam files) that support this study have been deposited to the NCBI Short Read Archive (SRA) under the project accession PRJNA660267.

Authors' contributions

M.B.-L designed experiments, performed experiments, analysed data and drafted the manuscript; V.V.-I. performed experiments and analysed data; C.R.A. analysed data; A.B.V.A. analysed data; A.G.S. performed experiments; M.S.-V. designed experiments; K.A., J.G.V, A.H., E.M., A.M.M., J.O., J.P.-P., K.S., J.K.W., G.Z. and A.LC. provided archaeological samples and provided archeological information. M.A.N.-C. designed experiments; K.H.I. analysed data; S.R. designed data analyses; M.C.Á.-A. conceived of the investigation, designed experiments, analysed data and drafted the manuscript. All authors contributed to and approved the final version of the manuscript

Competing interests

We declare we have no competing interests.

Funding

The project was funded by the Wellcome Trust Seed Award in Science 208934/Z/17/Z, and by project IA201219 PAPIIT-DGAPA-UNAM. We thank the PhD program of Biological Sciences at UNAM and CONACyT for providing funding to M.B.-L.