The next generation of genetic investigations into the Black Death

The study of ancient pathogens from human remains is as fascinating as it is essential for our understanding of past epidemics and human health through the centuries. However, ancient pathogens, such as Yersinia pestis (plague, Black Death) or Mycobacterium tuberculosis (tuberculosis), often do not leave unambiguous pathological lesions in the skeletal remains of their victims. Thus, it is not always straightforward to decide whether or not a long-dead individual was affected by a particular pathogen (1). For this reason, researchers have used ancient DNA (aDNA) techniques to search for genetic signatures of pathogens in human remains. In PNAS, Schuenemann et al. (2) turn this approach upside down. In their analysis of 99 human remains from a London burial site known to be associated with the Black Death epidemic of the mid-14th century (and of 10 control individuals not associated with the epidemic), they do not merely attempt to identify the etiological agent in these human remains but, rather, to reconstruct parts of its genetic code. Their aim is to identify whether changes in the genetic code of the Y. pestis-specific pPCP1 plasmid could be responsible for epidemiological differences between ancient and modern forms of Y. pestis infections. Such differences include symptoms, epidemiology, and time of year of peak mortality (2). Using high-throughput sequencing along with the latest DNA hybridization capture techniques, Schuenemann et al. (2) sequence the complete genome of the pPCP1 plasmid of Y. pestis along with some chromosomal markers and markers associated with the pMT plasmid. They find that the pPCP1 plasmid sequence in the ancient remains are consistent with previously sequenced modern variants and conclude that changes in the genetic code of the plasmid are therefore unlikely to be the cause of the known epidemiological differences between ancient and modern forms of the disease.

Challenges in Genetic Paleopathology

This study represents a contribution to a highly controversial field. Despite the relevance of genetic investigations into ancient epidemics, major breakthroughs have been rare and the field has been subject to recurrent skepticism regarding the authenticity of results (1, 3). In many cases, this skepticism was based on the lack of adherence to standard aDNA authenticity criteria (3). However, there is also the question of whether current molecular techniques are suitable to address the challenges particular to genetic paleopathology. One of these unique challenges is the fact that pathogens may have closely related and poorly characterized soil-dwelling sister species (3). Some of these species are known: Yersinia pseudotuberculosis, for example, is a soil-dwelling bacterium closely related to Y. pestis (4).

Schuenemann et al. provide a set of tools for sequence authentication in genetic paleopathology.

However, many more may not be known and may share genetic markers thought to be unique to the pathogen (1). The problem is further enhanced by the fact that DNA extracts from ancient bone and tissue samples are often swamped by environmental DNA from all sorts of microorganisms living on and in the bone. Studies of cave bears and Neanderthal remains, for example, have revealed that DNA of the organism the bone originates from often makes up less than 1% of the total DNA extracted, whereas more than 50% of the recovered DNA sequences are of unknown origin (5, 6). As a result, even strict adherence to aDNA authenticity criteria may result in erroneous conclusions if closely related species associated with the bone are mistaken for the pathogen. In particular, limitations of the PCR-based approaches for aDNA amplification and sequencing may facilitate the sequencing of modern contaminants rather than endogenous aDNA. PCR requires a certain minimum target molecule length to produce meaningful results. About 30 base pairs of informative sequence are needed to assign any given molecule confidently to its accurate coordinates within the genome (7). Thus, if each PCR primer is 20 bp long, molecules with a minimum length of 70 bp are required to obtain 30 base pairs of informative sequence (8). Identification using shorter PCR amplicons is possible but increases the risk for target misassignment. Previous studies, as well as this study, have shown that the average fragment length in ancient extracts can be significantly lower (2, 6, 9). Thus, PCR may only be targeting a small pool of fragments at the very top of the fragment length distribution. In this small pool of longer fragments, the percentage of modern contaminating molecules will be higher than among the shorter highly degraded ancient molecules. Thus, PCR will be biased toward modern contamination and may amplify contaminating molecules at the cost of endogenous molecules. This is particularly problematic if ancient human or human-associated bacterial DNA is to be amplified, because modern contamination can easily be introduced by handling the samples. Furthermore, because of the effort associated with sequencing long continuous stretches of highly fragmented DNA, genetic studies on ancient pathogens have often relied on sequencing very short marker sequences thought to be characteristic of the pathogen. The downside of this approach is that the shorter the obtained sequence, the harder it is to establish its authenticity.

High-Throughput Sequencing in Genetic Paleopathology

Schuenemann et al. (2) had to address these issues to demonstrate the authenticity of their sequence data and to underline the validity of their conclusions. The authors follow strict aDNA authenticity criteria (10, 11). Furthermore, by targeting the pPCP1 plasmid that is unique to Y. pestis, they reduce the risk for falsely identifying the pathogen. Initial screening of 99 human remains from a 14th-century Black Death victim burial site and of a control group of 10 individuals from an older non–Black Death-associated site was conducted using PCR-based approaches, and sequence data were obtained for standard chromosomal and plasmid markers. The true novelty of this study, however, lies in the techniques used to replicate and extend these initial results. For five Black Death victims and 5 individuals from the control group, the authors focus on sequencing the complete pPCP1 plasmid sequence of about 9,600 bp in length. Instead of PCR-amplifying the genomic target region, the authors use a recently published in-solution hybridization capture technique (12) to enrich aDNA extracts for the target region. This approach has several advantages over PCR-based approaches. It allows relatively easy sequencing of long continuous stretches of aDNA, making authentication of the obtained sequence easier. It also allows targeting fragments that are below the minimal amplification length of PCR. Thus, a fragment size range that likely has a smaller percentage of modern contaminating molecules can be targeted. Furthermore, hybridization capture combined with high-throughput sequencing allows for large numbers of individual template molecules to be sequenced. The result is a high coverage of the target region and a high confidence in the obtained sequence data. This high coverage allows the authors to evaluate their sequence data for damage patterns characteristic of aDNA sequences (13, 14). This, in particular, is one of the key improvements of the newly developed technology. By identifying aDNA-specific damage patterns, researchers can distinguish whether their sequence data are made up from ancient damaged or modern undamaged molecules, thus reducing the risk for mistaking modern contamination for genuinely ancient sequence data. As a control and reference for how these damage patterns should look in the 14th-century Y. pestis sequences, Schuenemann et al. (2) use the same hybridization capture approach to sequence the complete mitochondrial genomes of the same 10 individuals from which the Y. pestis pPCP1 plasmid was sequenced. This helps authentication of results in two ways. First, the mtDNA of victims should be more abundant than that of the pathogen, because not every cell gets infected by the bacterium (3). Thus, if the sequencing of the mitochondrial genome fails for an individual who has yielded Y. pestis DNA, or the amount of Y. pestis DNA greatly exceeds that of human mtDNA, this is a strong indicator that the Y. pestis DNA is not endogenous but is likely a modern contaminant. Second, because the pathogen DNA and the human mtDNA should be equally old, they should share similar damage patterns. Also, the approach helps in evaluating the negative controls. Controls that do not yield similar quantities of human mtDNA as the Black Death victims would be unlikely to yield Y. pestis DNA even if they were infected, and can therefore be considered poor controls.

Schuenemann et al. (2) find that in all sequenced individuals Y. pestis pPCP1 sequence reads are less abundant than the mitochondrial reads from the respective victim. The control individuals do not yield a single Y. pestis read, despite two of them having similar preservation as the best three Black Death victims. Crucially, this also confirms that no modern Y. pestis DNA used to produce the hybridization capture probes (12) was sequenced. The Y. pestis damage patterns are consistent with those of the respective human mtDNA. None of the Black Death victims yield sufficient pPCP1 reads to reconstruct the complete pPCP1 plasmid; thus, Schuenemann et al. (2) combine the pPCP1 reads from all individuals, arguing that all individuals are likely victims of the same strain of Y. pestis. This is certainly not ideal but is a perfectly legitimate approach, and there can be a good level of confidence in the authenticity of their results.

As is the case in many other aDNA studies, the study presented here was conducted on museum material excavated more than 20 y ago, and therefore almost certainly without later genetic analyses in mind. For studies on newly excavated material, the following approach may offer an additional level of authentication. If soil samples were collected along with the bone samples, the hybridization capture approach might allow for screening of these soil samples for the genetic signature of the pathogen in question. This would allow confirmation that the presumed pathogen DNA is indeed associated with bone samples rather than being characteristic of the environment in which the samples were found.

The most important aspect of this study is that it brings paleopathology into the era of high-throughput sequencing, an era that has already created a revolution in almost all other fields of aDNA research (8). It presents a technique suitable for obtaining long continuous stretches of sequence from ancient pathogens that may allow studying the genetics of ancient epidemics more easily in the future. By making use of high-throughput sequencing and hybridization capture technologies, Schuenemann et al. (2) also provide a set of tools for sequence authentication in genetic paleopathology, and thereby address one of the field's most sensitive problems. Future research using this approach may help to ease skepticism in the field and eventually lead to a better understanding of past epidemics.