Population studies have been performed analysing the hypervariable region of mitochondrial DNA, which accumulates nucleotide substitutions more rapidly than nuclear DNA because it lacks many of the DNA repair systems that operate in nuclear genes and because it is not under strong selective pressure. It is therefore the ideal marker to study the divergence between wild and domestic populations under the relatively short timescale of domestication. Moreover, since its mode of transmission is maternal, it does not undergo recombination and is therefore the perfect target to trace back maternal lineages and population migrations. When using a molecular phylogenetic approach, however, it is sometimes impossible to date these evolutionary and migration events with precision, since the approach is based on the assumption that the molecular clock is linear and universal [31], which does not always hold true (for reviews, see [33], [79] and [84]).

Analyses of the hypervariable region of mitochondrial DNA in modern populations of domesticated animals showed patterns that seemed to indicate the centres of domestication of the wild ancestors and subsequent migration routes of the domesticates [64]. For example, the data from a recent phylogenetic study of the mitochondrial hypervariable region in cattle from Southwest Asia, Africa and Europe showed a higher mitochondrial diversity in the former region and a dramatic reduction in this diversity in Europe and Africa [93]. Curiously, the diversity in Great Britain, i.e., at the endpoint of the migrations waves in which cattle are supposed to have spread from Southwest Asia into Europe, seems to be higher than in the core region of Europe. It is difficult to explain how the diversity in the periphery of the migration area can be higher than in the core region if one does not assume that the reduction in genetic diversity post-dated the dispersal of the species. The data from phylogenetic studies on extant cattle cannot be taken as conclusive proof that Southwest Asia was the centre of domestication during the Neolithic for several reasons. Firstly, if the data are to reflect the actual situation, studies of genetic diversity require a stringent and well thought outsampling protocol, which is difficult to follow rigorously. Secondly, calculations of the rates of sequence evolution depend on calibration of the date of bifurcation between the outgroup and the wild ancestor, a calibration that relies on the fossil record (e.g., in the case of cattle [20] and [65]). As pointed out by Dobney and Larson [27], however, these dates are not point estimates, but instead are ranges, thus widening the confidence interval of the mutation rate estimates. Furthermore, they depend on the time depth of relationships within the data set, where closely related species will yield faster mutation rates than distantly related species [51]. This will lead to an overestimate of the date of splits within species, a serious problem when dating recent events such as domestication [50]. Thirdly, these data may not reveal the timing and succession of the events that occurred between the time of domestication, i.e., roughly 10,000 years ago, and the present-day situation. In particular, the present-day population structure might result not only from the initial domestication and subsequent migration events, but also from breeding and selection procedures during historical and even very recent times [13] and [66]. Epidemic diseases such as the bovine pest that ravaged the European cattle populations from the 6th to the 18th century [43] might also have caused a dramatic reduction in genetic diversity. Finally, with these data, it is not possible to determine the time at which the loss of diversity occurred, and they are therefore not suitable for the identification of either the centre of domestication of the aurochs or the subsequent events leading to this dramatic loss of mitochondrial DNA diversity in Europe and Africa. To establish the nature of these events, it is necessary to complement both approaches – the archaeozoological approach and the molecular phylogenetic one – with palaeogenetic studies, which allow for internal calibration of the domestication process. The analysis of the mitochondrial hypervariable region of DNA preserved in Neolithic aurochs and cattle remains from the presumed centre of domestication, Southwest Asia, and all along the Neolithic migration routes into Europe, the continental route over the Balkans and along the Danube [12] as well as the Mediterranean route [103] and [106], and the areas where these migration currents finally mingled, in France [57], should shed light on the domestication and dispersal processes of cattle.

After death, bones do not always undergo total degradation but rather diagenesis, which transforms the initial inorganic and organic molecules and replaces most of the organic molecules by inorganic ones (fossilisation). During this process, the vast majority of DNA from the skeletal cells will be degraded. Under particular taphonomic (biological, physical, and chemical) conditions, however, DNA can escape total degradation and ‘survive’ for a long time. It is neither as yet known which factors (apart from temperature [91] and time), taphonomic agents and mechanisms are responsible for DNA preservation, nor whether there is an absolute time limit for DNA survival. Preservation of DNA in bones undergoing diagenesis is far from common (apart from that of permafrost samples), but is rather a lucky accident where several taphonomic factors and agents unite. This is the case in permafrost conditions where DNA is preserved at a higher frequency and quantity than under temperate conditions. The ancient DNA molecules are invariably heavily fragmented, chemically modified, and complexed with other molecules [34], [46], [74] and [92]. The escape from total degradation of these small DNA fragments is suspected to be favoured in particular microenvironments within the fossilising bone, so-called molecular preservation niches [34], where DNA is protected from enzymatic degradation either via adsorption to apatite within interconnected crystals [88] or in colloids via complexation with humic substances and minerals, rendering it unavailable for purification and enzymatic amplification [34], [36] and [38]. Such a preservation process implies a heterogeneous distribution of preserved DNA within the fossilising bone [35], which is indeed observed experimentally.

When comparing the quantity of DNA preserved in fossil bones that had been subjected to normal archaeological procedures, i.e., washing, drying and storage in rooms without air conditioning, to the quantities preserved in freshly excavated, untouched bone material, we have shown by analysing some 250 fossil bones from various burial contexts in zones ranging from temperate to arid (including Southwest Asia) that DNA is much more quickly degraded after excavation than during burial [82]. In fact, the success rate of PCR amplification from freshly excavated fossil bones was more than twice as high (46%) than that from fossils excavated using standard procedures (18%). Moreover, the average quantity of about 150-base-pair-long DNA fragments was roughly six times higher in freshly excavated bones than in ‘standard’ ones. Furthermore, fossils preserved in hot, arid regions may contain so little retrievable, endogenous DNA that it is degraded totally and quite quickly when the fossils are treated with standard excavation and post-excavation procedures, whereas genetic information can still be retrieved when they are analysed in their ‘dirt’ state directly after excavation [82]. Modelling of the quantitative data obtained showed that the post-excavation degradation rate of DNA in fossils preserved in moderate climatic regions, such as those in northern France, could be around 70 times faster than in the sediment. The degradation is likely due to depurination of the DNA, a chemical reaction following a first-order kinetic [59]. This post-excavation decay of 90% of the endogenous DNA molecules can be caused simply by an average storage temperature that is higher than the preservation temperature, combined with a decrease in the pH and a desalting effect due to the washing of the bone [82].

The consequence of this fast degradation of endogenous DNA once fossil bones are unearthed, washed, dried and stored is that they are more likely to produce false-positive results than bones preserved in permafrost or cave environments. In fact, the less authentic endogenous DNA a bone contains, the higher the risk that contaminating modern DNA from the environment will be analysed. Recent studies traced the pre-laboratory contamination of Neolithic human teeth to the DNA from the excavators, who had contaminated the samples during excavation, washing and subsequent anthropological and genetic studies [89]. This type of direct contamination of bones was also observed when 200- and 3300-year-old human bones were deliberately mishandled [18]. Further studies confirmed and extended this result to ancient dog specimens preserved in museums that all were found contaminated with human DNA [67]. The contamination process involves water that originates from the sweat of the excavators and experimenters [18] and/or from the standard washing procedures of fossil bones after excavation [89]. Contradicting results were obtained concerning the depth to which the contaminating DNA can penetrate in the bone: while one study showed that contamination occurs only on the outer 1–2 mm of the bones [18], another study demonstrated that handling of medieval human bones that had been excavated in the 1970s did not contaminate the bones with the experimenter's DNA [41]. Only DNA that was not specific to the experimenters was found and this was interpreted as stemming from the initial excavators and curators. An explanation for this observation is that buried fossilising bones and teeth might be very susceptible to contamination with environmental modern DNA during their excavation, when the pores of the calcified tissue are still open, whereas these collapse during post-excavation desiccation [41]. The contaminating exogenous DNA can show the same nucleotide base substitutions as ancient DNA due to subsequent base damage, and can therefore mimic ancient DNA, leading to erroneous results [89]. In contrast, other authors have claimed that contaminating DNA could be preferentially destroyed by bleach treatment of the bone powder [68]. In fact, these authors claim that contaminating DNA may be distinguished from authentic DNA based on the relative proportions of short and long fragments that can be amplified. This demonstration was not fully convincing, however, since the ratio between the quantities of supposedly authentic short to longer DNA fragments was not related to their absolute quantity, a correlation that would be expected if the interpretation were correct. The observed effect might rather have been caused by a combination of phenomena involving PCR inhibition, loss of DNA through bleach treatment, differences in the amplification efficiencies of the various primer sets, and/or carry-over or reagent contamination with modern DNA (see below).

To conclude, contamination of bones with environmental DNA occurs during handling, but may largely be prevented when bones are excavated and analysed with gloves and when washing is avoided.

The downside of the powerful method of PCR is its sensitivity to contamination with modern DNA molecules that are preferentially amplified. This is particularly problematic in the study of human remains, but also when domesticated animals, such as cattle and pigs, are studied because of the omnipresence of their DNA in our environment. These molecules are present in every laboratory, conveyed by the experimenters, and also present in the reagents, in particular, in deoxynucleotide preparations made from hydrolysed cattle or pig products [56], as well as in commonly used enzyme-stabilising proteins such as bovine serum albumin or gelatine (Champlot et al., in preparation. These modern contaminating molecules can easily be misinterpreted as authentic endogenous ancient DNA molecules in fossil samples where no or very little endogenous ancient DNA is preserved. This can lead to erroneous results reflecting the use of products of certain extant animal breeds for the production of reagents or the presence of certain human lineages amongst the employees of biotechnology companies involved in the production of the reagents, rather than the true past genetic diversity of aurochs, cattle, wild boars, pigs, and humans. For these reasons, not only the excavation and storage methods, but also the analytical ones have to be carefully monitored when working with fossil samples, in particular with human, bovine, and porcine samples that had been preserved in geographical areas and burial contexts in which the climatic and taphonomic conditions are unfavourable to DNA preservation. QPCR has proved to be the best method for careful quantification of both contamination by reagents and quantity of DNA found in fossil extracts. For our study of cattle domestication, we conducted a systematic quantitative analysis of reagent contamination to ensure the reliability of the results we obtained. We gamma-irradiated plastic ware and reaction water, and prepared large batches of reaction mixes. For each reaction mix, we established the frequency with which a mock positive amplification was observed and the quantities of molecules involved (we did not validate mixes that contained more than one molecule per ten reactions). We systematically sequenced a large number of mitochondrial DNA contaminants found in various batches of reagents over the years and observed that the contaminating sequences corresponded almost exclusively to the dominant European consensus sequence. Amplification from fossil extracts was considered reliable only when the frequency of occurrence of positive PCRs out of various amplifications was well above that seen with reagents alone. Such controls are critically required, especially when a sequence corresponding to the major European haplotype is obtained. Unfortunately, such rigorous monitoring was sadly lacking in a number of studies that reported bovine sequences identical to the major present-day haplotype, and hence the value of much of the literature reporting palaeogenetic studies of bovine (and porcine) domestication is limited.

Quantitative high-fidelity PCR systems help to increase the reliability of the production of DNA sequences from fossil extracts. This is a critical issue even for palaeogenomics [42], [53], [71] and [78], where the results obtained by large-scale sequencing need to be confirmed by PCR (see Hofreiter, this issue). When QPCR is routinely coupled with enzymatic fragmentation of DNA molecules produced in previous PCR and cloning steps, this can overcome one of the most severe threats to the authenticity of ancient DNA sequences, namely carry-over contamination. Since each PCR and each bacterial colony after cloning produces up to a billion copies of the original target molecule, concentrated in one tube, the danger of spreading these molecules is enormous. This is a major concern, since these molecules are completely identical to the original target molecule. It can distort in-house reproduction of previously retrieved data. Physical containment, positive air-pressure, UV-irradiation, and one-way circulation practices in the laboratory cannot be expected to decrease this contamination source to zero. We are routinely using a QPCR procedure with dUTP during PCR amplification, and preincubate each PCR with UNG to degrade potential contaminants from previous reactions. We have established that this UNG-coupled quantitative real-time PCR (UQPCR) eliminates 99.99% of DNA molecules from previous PCRs [81]. We also perform our cloning steps in an E. coli strain that incorporates dUTP in DNA, which allows us to degrade 94% of potential contaminant plasmid molecules [81].

By using UQPCR in conjunction with strict physical containment procedures, we add a considerable level of reliability to our analyses that cannot be attained with the strictest physical containment procedures alone, because no procedure can eradicate a major vector of contamination: the experimenter (see § 2.3.1).

When the two procedures, QPCR and UNG treatment, are combined in the so-called UQPCR [81], not only carry-over contamination is prevented, but also the error rate of nucleotide incorporation during PCR is considerably reduced. This is attributed to two phenomena. First, the enzyme uracil-N-gylcosylase (UNG) eliminates the majority of post-mortem nucleotide-base lesions occurring in ancient DNA, namely deaminated cytosines [52] and [92], thus preventing the major causes of erroneous copies of the original template. Second, we have established a strategy that most likely increases PCR fidelity. Indeed, fossil extracts contain not only inhibitors that decrease PCR efficiency, but also compounds that affect the fidelity of the Taq DNA polymerase, such as the ions Mn²⁺ [28], which are very often concentrated in fossilising bones and possibly co-purified bound to humic substances. We systematically quantify the inhibitory effect of each fossil extract on a dilution series of an internal DNA template and precisely dilute the extracts to minimise inhibition [80]. We believe that this also permits the dilution of the inhibitors that affect the fidelity of the reaction. Indeed, the percentage of non-synonymous base substitutions that we detected after sequencing of the resulting UQPCR products and clones amount to only 0.024% type-2 transitions and 0.046% type-1 transitions [83], whereas standard procedures report higher values: 0.18% and 0.11%, respectively [40], or 0.25 to 0.8% [88]. Therefore, this procedure considerably increases the fidelity of the amplification of ancient DNA molecules, yielding sequences that are more reliable.

An increase in PCR fidelity was also achieved via treatment of fossil extracts with hypochlorite [88]. Although this treatment was apparently relatively efficient in terms of fidelity, it was costly in terms of successful PCR amplification, since it caused the loss of 99% of the ancient DNA molecules present in the fossil bone [88]. In contrast, UQPCR achieves an even lower level of non-synonymous substitutions in the PCR products [83], while causing the loss of only a few molecules in only a subset of the fossil extracts (Champlot et al., in preparation).

In most palaeogenetic studies, the deduction of the authentic ancient DNA sequence is based on the evaluation of a certain number of clones. This number is crucial for the reliability of the deduced sequence. In 2005, Bower et al. [19] claimed that a minimum of 12 clones is necessary to deduce a correct consensus sequence from a fossil sample. This minimum number can rise to 30 clones per PCR product if the heterogeneity of the sequences is higher than average or if jumping PCR phenomena occur generating recombinant PCR products [21] and [104]. Using an in vitro assay, we were able to show, however, that direct sequencing of the PCR product is as powerful for the deduction of the consensus sequence as the proposed sequencing of 20 clones, because the majority sequence is directly read, while a 20% contamination with another DNA sequence can be easily detected [83]. Using these procedures on at least four PCR products per fossil sample, i.e., two independent PCRs from two different extracts of each sample, a high degree of reliability for the production of sequences from fossil extracts is guaranteed, so far equalled only by the procedure described by Stiller et al. [92]. To exclude post-mortem damage and polymerase misincorporations, these authors validated a sequence only if two lots of 12 clones obtained from two independent PCRs were 100% identical. Unfortunately, this seems to be the only study conducted according to these strict but very necessary rules. Instead, when the methods used lead to a high nucleotide misincorporation rate, consensus sequences are often determined in a somewhat arbitrary fashion from small heterogeneous sequence data sets, in which the chosen ‘consensus’ can sometimes differ from the majority of the clones found within some PCR products (e.g., [73]), or, even worse, differ from the sequence obtained directly from the PCR product [32].