Detection and Molecular Characterization of 9000-Year-Old Mycobacterium tuberculosis from a Neolithic Settlement in the Eastern Mediterranean

Background Mycobacterium tuberculosis is the principal etiologic agent of human tuberculosis. It has no environmental reservoir and is believed to have co-evolved with its host over millennia. This is supported by skeletal evidence of the disease in early humans, and inferred from M. tuberculosis genomic analysis. Direct examination of ancient human remains for M. tuberculosis biomarkers should aid our understanding of the nature of prehistoric tuberculosis and the host/pathogen relationship. Methodology/Principal Findings We used conventional PCR to examine bone samples with typical tuberculosis lesions from a woman and infant, who were buried together in the now submerged site of Atlit-Yam in the Eastern Mediterranean, dating from 9250-8160 years ago. Rigorous precautions were taken to prevent contamination, and independent centers were used to confirm authenticity of findings. DNA from five M tuberculosis genetic loci was detected and had characteristics consistent with extant genetic lineages. High performance liquid chromatography was used as an independent method of verification and it directly detected mycolic acid lipid biomarkers, specific for the M. tuberculosis complex. Conclusions/Significance Human tuberculosis was confirmed by morphological and molecular methods in a population living in one of the first villages with evidence of agriculture and animal domestication. The widespread use of animals was not a source of infection but may have supported a denser human population that facilitated transmission of the tubercle bacillus. The similarity of the M. tuberculosis genetic signature with those of today gives support to the theory of a long-term co-existence of host and pathogen.


Introduction
Tuberculosis is a major global cause of death and disease and around 2 billion people, about one third of the world's total population, are believed to be infected with tubercle bacilli [1]. However, only around 10% of infected persons become ill with active disease, and this high level of latent infection is an indication of long-term co-existence of human host and bacterial pathogen [2]. Tuberculosis is caused by a group of closely related bacterial species termed the Mycobacterium tuberculosis complex. Today the principal cause of human tuberculosis is Mycobacterium tuberculosis.
Mycobacterium bovis has a wider host range and is the main cause of tuberculosis in other animal species. Humans become infected by M. bovis, usually via milk, milk products or meat from an infected animal. It is estimated that in the pre-antibiotic era M. bovis was responsible for about 6% of tuberculosis deaths in humans [3,4]. Other members of the M. tuberculosis complex include the human pathogens Mycobacterium canettii, Mycobacterium africanum, and species usually associated with animal infections, such as Mycobacterium microti, Mycobacterium caprae and Mycobacterium pinnipedii.
Tuberculosis can cause characteristic skeletal changes, such as collapse of the vertebrae (Pott's disease), periosteal reactive lesions, and osteomyelitis [5]. Such paleopathological changes have been reported in pre-dynastic (3500-2650 BC) Egypt [6], and Neolithic (3200-2300 BC) Sweden culturally associated with the earliest cattle breeders [7]. These are the oldest cases of human tuberculosis confirmed by ancient DNA. Older cases recognized by skeletal changes alone were found in Neolithic Italy at the beginning of the fourth millennium BC [8,9].
Erosive lesions suggestive of tuberculosis have been found on fossil fauna from the Natural Trap Cave in Wyoming, dated from the 17,000 to 20,000 year level [10] and tuberculosis in one specimen was confirmed by biomolecular methods [11]. Initially it was believed that humans acquired tuberculosis from animals, especially after domestication [12][13][14]. Whole genome sequencing has since revealed that the M. tuberculosis complex has accumulated deletions over time, which can be used to distinguish individual species and lineages [15] and earlier ideas about the evolution of the M. tuberculosis complex have been revised [16,17]. An intriguing indication of the antiquity of the disease is the finding of non-specific morphological changes consistent with tuberculosis in a fossil Homo erectus dating from the middle Pleistocene (490-510,000 years BP) from Turkey [18].
The emergence of human infectious diseases has been linked to changes in human ecology and to interactions between populations [19]. The change from a gatherer-hunter lifestyle to settled farming communities appears to coincide with the appearance of diseases such as smallpox, measles, malaria, schistosomiasis, and tuberculosis [20]. Our aim was to investigate this stage of human history by the use of molecular methods to examine human remains that pre-date the earliest verified cases of tuberculosis, but with paleopathology consistent with this disease. A further aim was to elucidate the molecular characteristics of the causative organism. It is believed that the denser, settled populations associated with agriculture and animal domestication enabled human pathogens such as M. tuberculosis to be maintained indefinitely [21]. Therefore, we examined one of the earliest villages with evidence of both animal domestication and agriculture, Atlit-Yam [22], for the presence of tuberculosis in human remains with characteristic lesions. M. tuberculosis was confirmed in the skeletal remains of a woman and child, using both ancient DNA and bacterial cell-wall specific lipid markers. Deletion analysis indicates that the modern M. tuberculosis lineage characterized by the TbD1 deletion existed 9000 years ago.

Materials and Methods
The site of Atlit-Yam is now located 300-500 m offshore, (34u569 E, 32u42.59 N), 8-12 m below sea level in the North Bay of Atlit, 10 km south of Haifa ( Figure 1). Calibrated radiocarbon dates range from 9250-8160 years BP [23], indicating a date during the last phase of the Pre-Pottery Neolithic C period, when human society accomplished a full shift from hunting and gathering to farming, fishing and animal husbandry. The rich finds included botanical remains, tools, animal and human bones. The many animal remains that were excavated included goat (44%), cattle (43%), pig (9%), gazelle and deer (3.3%).
Human skeletons, which were embedded in dark clay, were carefully excavated and soaked in fresh-water tanks to dissolve the salts. The skeletal remains [24,25] were generally well-preserved (Supporting Figure S1A) and some showed paleopathological features consistent with a diagnosis of tuberculosis. Samples were taken for molecular examination from the skeletal remains of a woman buried together with an infant (Supporting Figure S1B). We analyzed the ribs, arm bones (adult) and long bones (infant). The work was done in separate centers to provide verification of data, and stringent precautions were taken against contamination (Supporting Materials and Methods S1). Separate areas and pipettes were used for extraction, PCR set-up and product analysis. Filter tips were used routinely and surfaces and equipment were cleaned before each assay. DNA extracts were prepared and, using the polymerase chain reaction (PCR), both multi-copy and single copy target loci were amplified (Table 1) and sequenced to confirm their identity. Screening PCRs detected the M. tuberculosis complex and nested or hemi-nested PCR was used to increase the likelihood of detection (Supporting Materiasls and Methods S1). A single copy conserved membrane protein locus (CMP) found in the M. tuberculosis complex was examined by PCR to assess the feasibility of seeking further single-copy loci. PCR target regions, based on specific deletions, were used to distinguish between M. tuberculosis and M. bovis. Extracts were analyzed further by reverse dot-blot hybridization of the M. tuberculosis complex-specific Direct Repeat (DR) region, a procedure known as spoligotyping [26]. In addition, samples from both the infant and adult were analyzed by high performance liquid chromatography (HPLC) for mycobacterial cell wall mycolic acids [27,28] (Supporting Tables S1, S2 and S3 and Supporting Figure S2 A and B). Long chain fatty acids were converted to pyrenebutyric acid-pentafluorobenzyl mycolates, and reverse phase HPLC examined for profiles similar to standard M. tuberculosis. Further normal and reverse phase HPLC was performed to give detailed profiles for each sample. These were used to determine the percentage ratios and absolute amounts of mycolic acids extracted from bone samples.

Paleopathology
The infant, though small in size, was estimated to be about 12 months old, based on crown development and long bone dimensions. On the inner aspect of the infant cranial bones were serpentine engravings (serpens endocrania symmetrica, SES; Figure 2A), a reliable diagnostic criterion for intra-thoracic inflammation [29] and associated with tuberculosis. The infant tubular bones also demonstrated lesions identified as hypertrophic osteoarthropathy (HOA), highly suggestive of tuberculosis [30] and characterized by the formation of an expanded shell of periosteal reactive bone ( Figure 2B and C). The woman was estimated to be around 25 years old, based on dental attrition, epiphyseal ring ankylosis and symphysis pubis. There was a slight periosteal reaction affecting the distal diaphysis of the one tibia available for examination, a bony change consistent with HOA [5,31,32]. However, the changes were not so marked as to be diagnostic.

Mycobacterium tuberculosis complex and M. tuberculosis DNA
M. tuberculosis complex DNA was detected in the bones of woman and infant (Table 2). Positive results with the multi-copy IS6110 [33,34] and IS1081 [35] PCRs were obtained with the rib sample from the woman and infant long bone sample, and confirmed by sequencing. An IS6110 123 bp product from the woman (right rib) and a 92 bp nested IS6110 product from the infant were obtained, identical to those in the NCBI database. Additionally, a 104 bp sequence identical to the relevant NCBI sequence in the IS1081 product was obtained from the infant.
The single copy TbD1 flanking PCR was positive from the infant sample and a complete DNA sequence for the 128 bp amplicon with the outer primers was obtained (Supporting Figure  S3A and B). The consensus sequence was identical to that in the NCBI database. The strong signal indicates the excellent preservation at this locus of the M. tuberculosis DNA template. Nested PCR was also successful. Weak positives were obtained with the outer primers from the female sample. These findings are evidence that the infecting organism was M. tuberculosis from a lineage in which the TbD1 deletion had occurred [16]. Results from the infant for the single copy CMP PCR were faint and a partial sequence was obtained (Supporting Figure S3C) with some mismatched bases compared with the database, attributed to poor

Lipid biomarkers of M. tuberculosis
Modifying an established protocol [28,36], long-chain fatty acids were extracted as pentafluorobenzyl (PFB) esters, and fractions corresponding to PFB mycolates were obtained (Supporting Tables S4, S5). After treatment with pyrenebutyric acid (PBA) these fractions produced PBA-PFB mycolates, which, after reverse phase HPLC, gave profiles closely similar to those produced by the M. tuberculosis complex, as indicated by a standard M. tuberculosis strain (Figure 3). Further normal and reverse phase HPLC gave detailed profiles for each sample, reinforcing the close identity with M. tuberculosis (Supporting Figure  S2 C and D).

Discussion
It is believed that inundation of the Atlit-Yam site occurred shortly after abandonment [24] and thereafter the environment remained unchanged for 9,000 years. The Atlit-Yam site was located within marshland; the graves were encased in clay, eventually covered by thick layer of sand and later by salt water, thus providing anaerobic conditions that retard degradation. The excellent preservation of the skeletal remains is consistent with the excellent physical state of the organic artifacts (wooden bowls, reed mats) that were found on the site. The paleopathogical lesions of SES in the infant cranium and HOA in the infant long bones and possibly also the tibia of the woman buried with the infant, presumed to be the mother, suggest that both suffered from, and died of, tuberculosis.
Anaerobic conditions are also conducive for DNA preservation [37] and DNA analysis supports the paleopathological diagnosis of tuberculosis. Overall, the PCR data provide strong evidence of the M. tuberculosis complex as specific DNA was detected in five different genetic loci, including the TbD1 locus with a deletion that is specific for a broadly-defined modern lineage of M. tuberculosis. Failure to detect human DNA (data not included) may reflect the greater stability of the GC-rich mycobacterial DNA, which additionally benefits from the robust hydrophobic bacterial cell wall [28,38,39].
Direct detection of cell wall mycolic acids specific for the Mycobacterium tuberculosis complex, without any amplification step, provides independent, robust confirmation of the presence of tuberculosis. The quantity of mycolic acids appeared lower in the infant sample (Supporting Table S5), in contrast to the DNA studies where the infant gave better results. However, the mycolate analyses were carried out on three combined rib samples from the baby, not all of which had been tested for MTB DNA. These extremely hydrophobic high molecular weight molecules are more stable than DNA and have been used previously to confirm diagnoses of ancient tuberculosis [27,28].
We conclude that both individuals in our study were infected with M. tuberculosis, and that our findings are supported when we consider the nature of the site, the stringent precautions taken to prevent cross-contamination and verification by the specific lipid biomarkers. Furthermore, we believe that this is the earliest report of the disease in humans that has been confirmed by molecular means. The infant is likely to have had disseminated primary tuberculosis: -the only DNA sequences for single copy loci were obtained from the infant material, which suggests a higher bacterial load during life. In infants less than a year old the present risk of developing active disease on infection with M. tuberculosis is as high as 43% [40] due to the inadequacy of their immune system. This compares with 5-10% in adults, 15% in adolescents, and 24% in children aged 1-5 years.
The size of the infant's bones, and the extent of the bony changes, suggest a case of acquired neonatal tuberculosis, in which an adult suffering from contagious pulmonary tuberculosis infects an infant shortly after birth. Childhood tuberculosis is closely linked to adult disease and is usually a sentinel event in the community, demonstrating recent transmission. Infant and maternal mortality rates from untreated tuberculosis in recent times was between 30% and 40% [41], so it is unsurprising for both mother and child to succumb and be buried together.
Spoligotyping should be a useful method of examining DNA from archaeological material as even fragmented DNA gives results, due to the increased sensitivity from the combination of amplification and hybridization [42]. M. tuberculosis complex DNA from the lesion of a 17,000-old extinct Pleistocene bison [11] yielded spoligotyping patterns most similar to Mycobacterium africanum or M. tuberculosis [43], and distinct from present day M. bovis. Zink et al. [44] obtained spoligotypes from ancient Egyptian   We carried out spoligotyping on specimens from the woman and infant but replicated typing gave inconsistent results, which suggests there may be poor DNA preservation of some of the single-copy spacer regions. The observed patterns do not match any in the International Data Base spoldb4: www.pasteurguadeloupe.fr/tb/spoldb4 but appear similar to an ancestral pattern. Results need to be interpreted with caution, as spoldb4 is based on data obtained from cultured organisms and spoligotyping has not been validated for application to DNA extracts prepared from degraded or archival specimens. The spoligotyping technique is based on modern M. tuberculosis strains from around the world. As the main variation in types is caused by unidirectional deletions, all ancestral strains are likely to produce a nearcomplete profile of the DR region and therefore to resemble each other. This may explain why the spoligotypes from the Atlit Yam skeletons resemble those of the Pleistocene bison [11].
Deletion analysis is a more robust method of examining ancient material [16,45], and based on the TbD1 deletion, the genetic lineage resembles modern lineages of M. tuberculosis [16,43,51].
Suggestions that human tuberculosis arose from M. bovis in hunted or domesticated animals have been revised since comparative genomic studies demonstrate that M. bovis represents a later lineage [16,17]. Members of the M. tuberculosis complex are genetically very similar and were believed to be the result of a clonal expansion following an evolutionary bottleneck 20,000-35,000 years ago [16,46,47]. However, further genomic studies of the M. tuberculosis complex indicate a more ancient origin of this group of closely related species than had previously been believed, and that possibly an early progenitor, perhaps similar to M. canettii, was present in East Africa as early as 3 million years ago [48,49]. The observation of non-specific lesions consistent with tuberculosis found in a 500,000 year-old skeleton of Homo erectus [18] may also indicate the long-term co-existence of host and pathogen, although the diagnosis in this particular case has been questioned. However, M. tuberculosis appears to have undergone long-term co-evolution with its human host prior to the evolutionary bottleneck and well before the development of agriculture and domestication, comparable to other long-term human pathogens such as Helicobacter pylori [50,51].
The present study of a population from 9250-8160 years ago, around the time of the first great transition from hunter-gatherers to a settled agriculture-based lifestyle [19], helps us to understand the nature of tuberculosis within the Middle East. Could the presence of cattle be pertinent? Atlit-Yam is among the very few Pre-Pottery Neolithic sites where domesticated cattle have been found. Furthermore, it is the only Neolithic site where there were quantities of bovine bones, indicating that cattle were a major dietary component [23]. We suggest that in the absence of detectable M. bovis, the cattle may be important by supporting a larger and denser human population, thus indirectly encouraging the conditions for the long-term maintenance and transmission of M. tuberculosis [21].

Author Contributions
Conceived and designed the experiments: IH HDD DEM GSB AMG EG CLG GKBG MS. Performed the experiments: IH HDD OYCL AMG EG VE CLG EL GKBG. Analyzed the data: IH HDD DEM GSB AMG EG VE EL GKBG MS. Contributed reagents/materials/analysis tools: GSB. Wrote the paper: IH HDD MS.