The past, present and future of ancient bacterial DNA

The birth of a field

The first breakthrough in aDNA occurred when Higuchi et al. [6] were able to extract a small genetic fragment from a 150-year-old museum specimen of the extinct horse relative quagga through laborious bacterial cloning. Paleogenetics was soon propelled by the development of the polymerase chain reaction (PCR) [7]. PCR produces abundant genetic material from infinitesimal amounts of aDNA given a primer to guide the amplification. With the choice of primer, researchers can decide whether they want to take a targeted approach to generating aDNA from ideally a single species (e.g. amplifying a gene of a specific toxin), or an untargeted approach generating aDNA from multiple species (e.g. amplifying 16S rDNA found in all bacteria). The success of amplification on human, animal and plant remains spurred the first bacterial aDNA studies targeting Mycobacterium tuberculosis [8] and Mycobacterium leprae [9]. Müller and colleagues [10], however, highlight that the presence of mycobacteria from the human microbiome or soil could lead to false positives without further scrutiny, casting doubt on the validity of early findings lacking subsequent authentication. Not only did the extraction of mycobacteria prove difficult, but the whole field quickly reached limits imposed by the then available technology because PCR is far from optimal for analysing aDNA. Successful amplification requires relatively long, well-preserved, sequences that cannot easily be acquired from very short degraded ancient samples [11] (typically shorter than 100 base-pairs [12]). Additionally, primers for aDNA amplification must be specifically tailored to ancient targets, which poses unique challenges in design. Using contemporary sequences, a section of DNA must be identified that is taxon-specific and short enough to be covered by short aDNA fragments, but also long enough to be used for amplification. Success is not guaranteed, given that mutations can occur that obstruct the design of viable primers for ancient material from modern-day counterparts. The amplification of minuscule ancient genetic material required for detection, coupled with difficulties of authentication, led to widespread contamination from modern-day DNA [13]. Due to the large number of PCR cycles needed for ancient samples, the procedure is liable to amplify trace amounts of modern-day and environmental DNA or that of postmortem colonizers [14]. Additionally, degradation of DNA over time can lead to sequencing errors because enzymes used for amplification preferentially bind to non-damaged modern contaminants [15]. The debate concerning the authenticity of early aDNA studies led to the development of rigorous validation processes that are still used today [12].

The advent of high-throughput sequencing

The advent of high-throughput sequencing (HTS) rejuvenated the field of paleomicrobiology as sample and data output increased by orders of magnitude [12] and brought new possibilities for authentication. The mass of sequence data led to a growth spurt in the field of computational sequence analysis, allowing in silico verification based on sequence patterns [16]. Signatures of postmortem damage (PMD) create nucleotide misincorporations characteristic of aDNA [17]. Generally, two forms of characteristic mechanisms of PMD are used for verification: depurination and cytosine deamination [18]. In depurination purine bases are removed through the cleavage of a N-glycosyl bond, leaving an abasic site that is susceptible to subsequent β-elimination, which breaks the DNA backbone [19]. This fragmentation causes an overrepresentation of purine bases preceding strand breaks, which is a pattern unique to ancient DNA. Fragmentation also results in sequences that are typically shorter than 100 bp [20], which is additionally used to differentiate ancient sequences from modern contaminants harbouring longer sequences [12]. In cytosine deamination, cytosine loses an amine group and is thus converted to uracil, which is registered as thymine in sequencing [20]. Bases become predominantly deaminated towards the exposed single-stranded ends of fragments, leading to an increase of misincorporations towards the 5ʹ end, another distinctive signature used for authenticating aDNA [21]. Age is not the only contributing element in PMD, because factors such as temperature, depositional conditions, post-excavation handling and source tissue can also influence the signatures, so that contaminants may also exhibit damage patterns [21]. However, it was shown that younger (50–100 years old) samples exhibited little PMD and a linear correlation of damage with time was discovered [22]. Misincorporations and fragmentation can be detected through alignment of reads to extant bacterial genomes, where the removal of PCR duplicates and high coverage ensure that misincorporations can be distinguished from genuine substitutions. Provided there is a genome available for the organism of interest, alignment allows the depletion of modern contaminants lacking PMD signatures [23].

The development of targeted capture sequencing enabled enrichment using modern DNA or RNA as bait, making the recovery of ancient genetic material more financially viable [24–26]. Enrichment facilitates the sequencing of miniscule amounts of ancient material from a species of interest amidst within-sample (endogenous) contaminating sequences from other species and environmental (exogenous) modern sequences. Targeted enrichment led to the first ancient bacterial genome of Yersinia pestis , which was recovered from the teeth of a 14th-century plague victim at 30-fold coverage [27]. The isolation of a high-coverage Y. pestis paleogenome verified earlier work by Drancourt et al. [28]. This confirmation put to rest the controversy around the possibility of isolation of the plague bacterium from teeth [29], which was sparked by the negative findings of Gilbert et al. [20] and opened up the possibility of a comparative genomic approach using aDNA.

Comparisons using paleogenomes facilitate a better understanding of evolutionary processes through time, enabling the inference of ancient demography and admixture, within-population evolution and calibration of the molecular clock [12]. In clock calibration the evolutionary rate is quantified, which describes genetic change as a function of time and in turn enables the dating of evolutionary events [30]. For accurate rate estimation, chronological anchors need to be supplied, which is especially fraught with difficulty for bacteria, since fossil evidence is nonexistent [31]. Further uncertainty is contributed by confounding factors such as the dependence of rate estimates upon the time spanned by sampling. Rates estimated by comparing species that have diverged millions of years ago are orders of magnitude slower than rates estimated within lineages that have diverged over decades [32]. This temporal inconsistency was named the time-dependent rate phenomenon and leads to discord concerning the timing of evolutionary events, because the interpolation of rates across different time periods is contentious [33]. Reported bacterial rate estimates reveal a broad spectrum between 10 substitutions per megabase per year (Mby⁻¹) in Neisseria gonorrhoeae [34] and 0.001 Mby⁻¹ in M. tuberculosis [35]. Chronologically, aDNA provides sequences dating within the gap that spans rate estimates drawn from comparison between and within species [33]. Therefore, datasets that comprise contemporary and ancient samples are apt to address the conflict in reported evolutionary rates. Bos and colleagues [36] investigated the impact of the inclusion of aDNA of increasingly larger time periods using three different Y. pestis datasets. The estimated mean ages were pushed further back the older the included sequences were and showed that the inclusion of Late Neolithic and Bronze Age samples decreased dating uncertainty immensely. Duchêne et al. [37] analysed aDNA from three different species ( M. leprae from Schuenemann et al. [38], M. tuberculosis from Bos et al. [39] and Y. pestis from Wagner et al. [40]) spanning approximately 2000 years and confirmed the existence of a rate phenomenon. Where applicable, the discord between rate estimates spanning different time periods will be detailed in the following, as it provides an important application of aDNA in bacterial genomics, a field particularly impacted on by incongruity.

Alongside high-profile extraction like Neandertal [41] and woolly mammoth [42] DNA, bacterial aDNA research grew steadiliy (Fig. 1) from the study of 16S rDNA of a single species to the recovery and comparison of full genomes from multiple genera. The history of bacterial aDNA research (Fig. 1) has led from the study of 16S rDNA of a single species to the recovery and comparison of full genomes from multiple genera. The key methodological stepping stones for paleogenetics are the development of PCR and HTS, leading to key publications such as the first amplification of bacterial aDNA [8] and the first retrieval of a full bacterial paleogenome [27]. Duchêne and colleagues’ study [37] is also an important achievement, where aDNA was first used across a range of species to make more comprehensive statements about the history of the bacterial domain. Also vital for analysing the bacterial past is the discovery of a growing collection of viable sources for the retrieval of aDNA, each with its distinct opportunities and obstacles.

Sources of bacterial aDNA

The beginning of sequencing, and therefore also paleogenetics, was facilitated by targeted PCR amplification, making the identification of candidate targets from ancient samples vital to the success of early attempts at acquiring bacterial aDNA. Seminal research was directed at the recovery of genetic material from bone as aDNA extraction is destructive and often unsuccessful. Bones are the most common of the invaluable and finite archaeological resources suited for aDNA retrieval. The choice of study organism fell upon pathogens causing diseases that can occasionally be identified by skeletal lesions, to further increase chances of successful extraction. The choice fell upon tuberculosis (TB) and leprosy, leading to the first retrieval of bacterial aDNA of M. tuberculosis by Spigelman et al. [8], followed by Rafi et al. [9] amplifying M. leprae . The realization that aDNA can be recovered from bone tissue without visual evidence of disease led to an increase in available sequences, facilitating epidemiological investigations of populations [43] and mixed infections [44].

The extraction of aDNA from M. tuberculosis soon moved from bones to mummies, where natural mummification can lead to preservation, leaving bacterial DNA sufficiently intact for amplification [45]. Salo et al. [46] were able to acquire bacterial DNA from a 1000–2000 year-old mummified Peruvian body. The study revealed pre-Columbian presence of TB in the Americas, a disease formerly believed to be introduced by explorers [39]. Another successful extraction of bacterial aDNA from mummies was conducted by Cano et al. [47] through amplification of 16S ribosomal DNA, genes common to all bacteria, providing the first glimpse into ancient microbiomes. The subject was the 5300-year-old exceptionally well-preserved Tyrolean iceman. Similarities to the modern human gut flora were found with members of the genus Clostridium . The concern around PCR amplification producing false positives makes the validity of these early findings questionable, especially because Cano and colleagues also found clostridia in the grass sample taken from the surroundings of the iceman.

Donoghue et al. [48] were able to add calcified pleura to the sources from which M. tuberculosis could be amplified with PCR. Furthermore, historic medical specimens, such as organs embedded in paraffin, were proven to be viable subjects for paleomicrobiology. Jackson et al. [49] amplified Bacillus anthracis DNA from anthrax victims from 1979. Devault et al. [50] later used the alcohol-preserved colon of an 1849 cholera victim to generate the first and only paleogenome of Vibrio cholerae .

The inventory of aDNA sources was broadened by Drancourt et al. [28], who realized that teeth are both easily retrievable from burial sites and highly vascularized and thus a resource for bloodborne diseases [51]. Teeth as a preserving molecular niche minimize the potential for contamination through postmortem colonization of bacteria [52]. Drancourt et al. were able to retrieve DNA from Y. pestis , and thus found the aetiological agent for the black death. They suggest that dental pulp, as a better protected niche than bones that is equally expendable, should be used to detect ancient pathogens.

Early claims of isolation of bacterial aDNA from the environment were made by Priscu et al. [53] and Christner et al. [54] through the amplification of 16S rDNA from meltwater and ice from Lake Vostok, Antarctica. Willerslev et al. [55], however, pointed out that one of the organisms identified is a common laboratory contaminant and the only species recovered in both studies was also found in PCR controls. Further doubt was cast by subsequent studies that were unable to recover bacterial DNA from Lake Vostok ice [56]. Willerslev et al. [57] addressed their own criticism and amplified 16S rDNA under stringent laboratory conditions. Their samples were frozen under optimal conditions in Antarctica and northern Siberia.

Besides mummified remains, ancient human microbiomes commonly endure only in two substrates: coprolites and dental calculus [58]. Coprolites are mummified or petrified faeces that can contain the composition of intestinal microbiota at the point of death, allowing the study of the gut microbiome [59]. As myriads of micro-organisms are active in faeces, decomposition is rapid, leading to quick sample degradation [58]. Although found in dry, cold and tropical environments [60, 61], preservation is only optimal under extremely dry or frozen conditions [47]. The detection of coprolites is difficult due to partial fossilization or fragmentary or amorphous preservation [62]. Despite their rarity, Tito et al. [63] were able to use HTS on coprolites from 1300-year-old Mexican cave samples. Their findings suggest that human microbiomes were previously more geographically structured. Apart from bacteria, the parasitological composition can be assessed through the presence of eggs, larval protein and parasite DNA [58], while diet can be assessed through plant, fungal and faunal remains [64]. A combined analysis of diet, microbiota and parasites allows for estimates of ancient dietary habits, lifestyles and cultures [65]. The comparison of microbiomes via coprolites has revealed a similarity between ancient and extant rural samples that indicates a compositional shift of microbiota in the modern cosmopolitan lifestyle [66].

The third possible source for human microbiomes is dental calculus. Calculus consists of dental plaque, saliva and crevicular fluid forming calcified bacterial biofilms incrementally through deposition and mineralization [58, 67]. Composed of layered structures, calculus produces a serial record of the oral microbiome showing the individual’s life history of diet and disease [67, 68]. DNA preserves well in calculus when compared to other sources, with preliminary microscopy analyses revealing whole bacteria conserved as rods and cocci [69]. Preus et al. [70] first proved the existence of DNA in calculus through transmission electron microscopy, which led Fuente et al. [71] to use species-specific PCR primers on 16S rDNA from five bacterial species. Adler et al. [72] were the earliest to assess the full microbiome by using 16S rDNA primers. Comparison to modern samples showed a shift to a disease-associated composition of oral microbiomes during the Industrial Revolution, with dental caries and periodontal disease-linked bacteria Streptococcus mutans and Porphyromonas gingivalis . Warinner et al. [67] used HTS to reveal key virulence and antibiotic resistance genes present in a medieval German specimen. Quantifying DNA preparation showed that the extraction of 1000 times more aDNA was possible from calculus compared to bone from the same individual. The abundance of extractable aDNA combined with the dominance of bacteria within the sample makes calculus a rich resource for paleomicrobiology [67]. However useful, using calculus as a source also has limitations. Calculus limits species of interest to members of the human oral microbiome and possibly provides a skewed representation of the past microbial diversity. Through the comparison of modern dental plaque and modern dental calculus, Velsko et al. [73] showed that a systematic bias in microbial composition is introduced in the calcification process, calling the inferences of Adler et al. into question.

Looking at all viable source of bacterial aDNA (Fig. 2), the initial study of Drancourt and colleagues should be highlighted as having discovered a valuable resource for human bloodborne pathogens. Dental pulp is better protected from contamination than bone, more easily available than coprolites and not as skewed in composition as calculus. Seminal studies in ice cores [53–57] should also be highlighted as the only source making ancient environmental bacteria accessible. Ice cores additionally have superior temporal reach compared to other sources [74], which is potentially crucial for resolving the discord in bacterial rate estimates and giving a glimpse of evolutionary dynamics outside a constrained host niche. Besides considerations of the source material for bacterial aDNA studies, the field faces systematic challenges unique to microbial research within archaeogenetics.

Challenges in bacterial aDNA studies

Archaeological samples used in bacterial aDNA studies show a mixture of endogenous (ante-mortem) viable ancient material and exogenous (postmortem) sequences from colonization after death of the organism [12]. Endogenous DNA is made up of host-associated commensal taxa and epidemic pathogens [12]. Exogenous DNA contains environmental bacteria, contamination from human interaction with samples, suboptimal storage and laboratory contaminants [12]. Bacterial aDNA retrieval is prone to false positives, given that many pathogenic bacteria are members of the same genera as environmental bacteria [75]. Authentication via PMD can be thwarted by postmortem colonization at different ages, resulting in a range of damage patterns from ancient to extant that impedes distinction [76]. A key methodological advancement for overcoming the challenge of contamination was the development of mapDamage [77], which is a computational tool that analyses PMD signatures and allows the identification of aDNA within a chronologically heterogeneous genetic sample.

Other than false positives, assessing bacterial communities can be inherently problematic in paleomicrobiology. The equilibrium of the microbiome can shift quickly after death, with anaerobic bacteria persisting more successfully, which can lead to false conclusions about ancient microbiomes [78]. Bacterial aDNA studies are further altered by the differential conservation of bacterial DNA. GC content and the composition of the cellular membrane influences preservation, so that GC-rich Gram-positive bacteria persist best [65, 78]. Membranes rich in mycolic acid allow for better protection against enzymatic degradation, as is the case in mycobacteria [79]. Ziesemer et al. [80] highlighted that no study has systematically investigated how age-related degradation might shift the reconstruction of ancient microbial communities. Heeding their own advice, Ziesemer et al. looked at systematic amplification bias of the V3 regions in 16S rDNA. They showed more successful amplification for shorter V3 regions, which might have skewed earlier, untargeted aDNA studies.

Despite the inherent challenges they pose, some characteristics of bacteria favour the study of paleogenomes. The bacterial bauplan harbours unique benefits for preservation, such as increased resilience based on membrane composition. Glycosidic ether lipids and hopanoids are components of bacterial membranes not found in archaea or eukaryotes, which protect DNA from enzymatic degradation [81]. Intracellular DNA protected in dormant or dead microbes conserves better than naked DNA [65]. Aided by benefits in preservation, and despite challenges unique to bacteria, researchers have been able to generate a tremendous amount of insight from the study of bacterial aDNA.

Human pathogens

The development of agriculture brought food security and a settled lifestyle to human societies, but also led to an increase in infectious disease and is considered to be the origin of many important infections (see e.g. Harper et al. [82]). The progression to denser, more fertil,e human populations in proximity to domesticated animals was coined the Neolithic demographic transition [83]. Molecular dating has hitherto linked the emergence of at least three pathogens pivotal for paleogenetics with the Neolithic transition: Y. pestis, M. tuberculosis and M. leprae [84].

Human microbiome

Paleomicrobiology has not stopped at the recovery of individual genomes, but has also assessed ancient microbiomes. Humans carry microbes comprising thousands of different species that collectively outnumber our own cells by at least an order of magnitude [154, 155]. The longstanding interaction of these co-inhabitants of our bodies has led to a growing appreciation of the human microbiome in studying human evolution. The focus of research has shifted away from individual pathogenic microbes towards an investigation of microbiomes as mutualistic components of metabolism, immune development and neurological function [155]. Ancient microbiome studies have drawn primarily from three sources– mummified remains, coprolites and dental calculus – and can be separated into studies focusing on oral or gut microbiomes [156].

The evolution of the oral human microbiome can be assessed through dental calculus and is primarily shaped by two major shifts in human diet: a shift towards carbohydrate-rich diets in the Neolithic demographic transition (~10 000 YBP) and the arrival of industrially processed flour and sugar (ca. 1850) [72]. Adler et al. [72] used HTS on dental calculus from 34 teeth dating from before the development of agriculture to the mediaeval period, showing that abandoning the hunter–gatherer lifestyle led to a disease-associated reconfiguration of the mouth microbiome. Warinner et al. [67] analysed dental calculus from the teeth of four German individuals from 950 to 1200 CE showing mild to severe periodontal disease. The sample was dominated by the so-called red complex pathogens, a group comprising Tannerella forsythia , Porphyromonas gingivalis and Treponema denticola , which supply the most important pathogens in adult periodontal disease [157]. Besides long-term carriage of opportunistic pathogens, Warinner et al. reported the presence of genes homologous to putative antimicrobial resistance genes. Weyrich et al. [158] analysed dental calculus of five Neanderthal specimens, a historic wild-caught chimpanzee and a modern human. The oral microbiome of the Neanderthal was closer in composition to that of the historical chimpanzee than the microbiome of modern humans. Pathogens associated with caries, causing the decay of teeth, such as S. mutans and members of the red complex were found in the Neanderthal samples [158]. The paleogenomes of Campylobacter gracilis, Propionibacterium propionicum , Fretibacterium fastidiosum , Eubacterium infirmum , Peptostreptococcus stomatis and Eubacterium sphenum were retrieved from the Neanderthal extractions.

The evolution of the gut microbiome can be assessed from both coprolites and mummified organs, which was successful early in paleomicrobiology. Ubaldi et al. [45] paved the way by proving that 16S rDNA of multiple species can be amplified from 900 to 1000 year-old Andean mummies. Ubaldi et al. predominantly discovered bacteria of the genus Clostridium . Cano et al. [47] amplified 16S rDNA of a broad range of bacteria from the 5300-year-old Tyrolean iceman mummy. The stomach was entirely composed of Burkholderia picketii, an aquatic bacterium possibly originating from meltwater, entering the body. The colon harboured several Clostridium species characteristic of the contemporary human faecal flora. Rollo et al. [73] conducted a study analysing the persistence of bacterial DNA, comparing the Tyrolean iceman to a mummy dated to 1918. Rollo and colleagues showed that the characteristic composition of extant microbiota is still present in the younger mummy. However, even under almost ideal conditions of preservation, the 5300 year old sample is almost entirely composed of Clostridium. Clostridium species are likely to persist better as they are endospore-forming, low-GC, Gram-positive bacteria, which are all qualities previously described as beneficial for preservation by Willerslev et al. [57]. Anaerobic species like Clostridium also survive due to a rapid decrease in oxygen concentration postmortem, indicating that the uncovered abundance of Clostrdium is unlikely to reflect the true state of the microbiome at the point of death [73]. Luciani et al. [159], upon analysis of the same Andean mummy as Ubaldi et al., found the pathogen Haemophylus parainfluenzae. Santiago-Rodriguez et al. [160] contributed the third study on the Andean mummy using HTS to characterize the microbiome, finding putative antimicrobial resistance genes. They also found an abundance of Clostridium , which comprised up to 96.2 % of the species in the gut. Lugli et al. [161] reanalysed the Tyrolean iceman using HTS data generated by Maixner et al. [148] and predominantly found Clostridium and Pseudomonas . Lugli et al. were also able to reconstruct five paleogenomes of the species Clostridium perfringens, Clostridium algidicarnis, Pseudomonas veronii and Pseudomonas fluorescens and a genome of the genus Clostridium that could not be assigned a species. The frequent retrieval of clostridial species could be attributed to their better preservation. However, many members of the genus are also found in soil [162]. Retrieval of clostridia from ancient samples could be false positives resulting from sequence similarity with soil-dwelling relatives that colonize postmortem and subsequently acquire PMD sequence patterns. Although bacteria colonizing after death can be genuinely ancient, they are misleading concerning past microbiota and the humans that harboured them. The continual reuse of mummified samples in aDNA gut microbiome research shows how difficult the acquisition of viable specimens is. Fortunately for paleomicrobiology, coprolites are another source for gut microbiomes.

Tito et al. [63] first analysed two coprolite samples dated to 1300 YBP using HTS and found a dominance of the phyla Bacteroidetes , Firmicutes , Actinobacteria and Proteobacteria typical of contemporary faecal microbiomes. The two ancient samples were more genetically alike than modern samples, possibly owing to biogeographical structuring. Tito et al. suggest that microbiomes were formerly more geographically structured, possibly due to more locally restricted diets. Cano et al. [59], through the analysis of coprolites dated to 180–600 CE, showed that the extinct Amazonian Huecoid and Saladoid cultures could be distinguished based on their coprolites. The difference possibly stems from host genetics or differences in diet [59]. Tito et al. [66] conducted the most extensive study to date, incorporating coprolites from three archaeological sites ranging from 8000 to 1400 YBP. A source analysis showed that two of the sites were either of unknown source or more alike to compost than human gut. The remaining Mexican sample, which featured in their previous study [63], showed higher similarity to contemporary rural microbiomes than urban communities. A modern lifestyle may have influenced ancestral mutualistic ties between humans and bacteria, possibly increasing the risk of autoimmune diseases amongst other conditions [63].

The study of Weyrich and colleagues [158] shows how rich of a resource calculus is, having constructed paleogenomes from six different species. The importance of the study also lies in the age of the paleogenomes, which date to 48 000 YBP, and are thus by far the oldest bacterial genomes, with the next most ancient paleogenomes dating to 5300 YBP [161]. The temporal reach of dental calculus is only surpassed by ice cores, which are the only source of environmental bacteria.

Non-human associated bacteria

Most environmental aDNA comes from permafrost due to the excellent conditions for preservation. Whether aDNA yielded from ice can be compared with other aDNA samples is questionable, as microbes could potentially remain metabolically active within micro-encasings of liquid water [74]. Active bacteria could potentially repair their DNA, influencing PMD signatures [74]. Frozen samples of potentially active bacteria could be a reservoir of adaptive potential upon thawing of the ice, given the microbial propensity to transfer genetic material horizontally. Bidle and colleagues [74] point out that the pace of evolution after major global glaciations seems to be increased. The first environmental samples of bacterial aDNA were provided by Priscu et al. [53] from an ice layer dated to 1 million YBP from Lake Vostok. Christner et al. [54] amplified 16S rDNA in five 20 000-year-old ice cores and found most bacteria to be related to Bacillus and Actinomycetes . Willerslev et al. [57] used 16S rDNA amplification on ice dated up to 800 000 YBP and found it dominated by Actinobacteria . Bidle and colleagues [74] used 16S rDNA amplification on Antarctic ice belonging to the oldest known samples reaching up to approximately 8 million YBP. The ice harboured Alphaproteobacteria , Acidobacteria , Firmicutes and Cytophaga–Flavobacterium–Bacteroides divisions. D’Costa et al. [163] investigated 30 000-year-old Beringian permafrost to look at genetic content through HTS. Genes coding for the resistance to beta-lactam, tetracycline and glycopeptide antibiotics were found, revealing ancient evidence of antimicrobial resistance. Segawa et al. [164] collected ice core samples from the Kyrgyz Republic dated to 12 500 YBP to look at the mutation rates of cyanobacteria through comparison to modern samples. Looking at the internal transcribed spacer region of two cyanobacterial operational taxonomic units, Segawa et al. found rates of 10 Mby⁻¹.

Bidle and colleagues’ work [74] is exceptional within paleogenetics, as it provided the oldest genetic evidence of bacteria at 8 million YBP. The study of environmental aDNA has thus spanned larger time periods than was ever possible by analysing human remains, although no bacterial paleogenome has been reconstructed from environmental aDNA to date, underlining the inherent challenges at this frontier of ancient DNA research.