Spiking of Contemporary Human Template DNA with Ancient DNA Extracts Induces Mutations Under PCR and Generates Nonauthentic Mitochondrial Se
http://www.100md.com
分子生物学进展 2004年第5期
* Institute of Anthropology and Human Genetics, Division of Molecular Genetics, University of Tübingen, Tübingen, Germany
Department of Zoology, Natural History Museums and Botanical Garden, University of Oslo, Oslo, Norway
E-mail: bachmann@nhm.uio.no.
Abstract
Proof of authenticity is the greatest challenge in palaeogenetic research, and many safeguards have become standard routine in laboratories specialized on ancient DNA research. Here we describe an as-yet unknown source of artifacts that will require special attention in the future. We show that ancient DNA extracts on their own can have an inhibitory and mutagenic effect under PCR. We have spiked PCR reactions including known human test DNA with 14 selected ancient DNA extracts from human and nonhuman sources. We find that the ancient DNA extracts inhibit the amplification of large fragments to different degrees, suggesting that the usual control against contaminations, i.e., the absence of long amplifiable fragments, is not sufficient. But even more important, we find that the extracts induce mutations in a nonrandom fashion. We have amplified a 148-bp stretch of the mitochondrial HVRI from contemporary human template DNA in spiked PCR reactions. Subsequent analysis of 547 sequences from cloned amplicons revealed that the vast majority (76.97%) differed from the correct sequence by single nucleotide substitutions and/or indels. In total, 34 positions of a 103-bp alignment are affected, and most mutations occur repeatedly in independent PCR amplifications. Several of the induced mutations occur at positions that have previously been detected in studies of ancient hominid sequences, including the Neandertal sequences. Our data imply that PCR-induced mutations are likely to be an intrinsic and general problem of PCR amplifications of ancient templates. Therefore, ancient DNA sequences should be considered with caution, at least as long as the molecular basis for the extract-induced mutations is not understood.
Key Words: ancient DNA extracts ? mitochondrial DNA ? mutation ? PCR errors ? PCR inhibition
Introduction
The retrieval of nucleotide sequences from ancient DNA templates requires technical skills and extensive expertise. The major shortcoming of ancient DNA research is that authentic sequences can only be determined in an indirect way. This is particularly true for samples from extinct taxa, i.e., when no high-quality DNA sample is available. To overcome the problem, short overlapping PCR products are cloned, and the authentic sequence is subsequently deduced as the consensus of many individual sequences that are derived from several independent amplifications (Cooper and Poinar, 2000; Hofreiter et al. 2001a,b). Such cloned PCR products are usually heterogeneous due to contamination with contemporary DNA, PCR artifacts, and assumed postmortem damage of authentic DNA molecules.
Contamination with modern DNA has plagued ancient DNA research, and this is a matter of particular concern when working with hominid samples. It might be impossible to prove authenticity of the obtained data simply because of the expected very high level of sequence similarity between authentic ancient DNA from a hominid fossil and contaminating contemporary DNA. Even minute amounts of contaminating DNA might question the results obtained by amplification of ancient DNA using the extremely sensitive polymerase chain reaction (PCR). However, it must be stressed that contamination is a general problem in ancient DNA research and affects the analysis of samples from any taxon (Pusch et al. 2000). Usually, PCR protocols are optimized using DNA from closely related taxa and carryover contaminations might occur.
PCR artifacts such as PCR jumping, allelic dropouts, regular polymerase errors, and assumed postmortem damage of authentic DNA molecules are also great obstacles in the retrieval of authentic sequence data from ancient DNA (P??bo 1989; Willerslev et al. 1999; Hansen et al. 2001; Gilbert et al. 2003a). Strand breaks decrease the average molecular weight of nucleic acids (P??bo 1989). Oxidative and hydrolytic damage may lead to incorporation of incorrect nucleotides or block strand elongation during PCR (Willerslev et al. 1999). The conversion of adenine to hypoxanthine and cytosine to uracil, respectively, are considered common forms of DNA damage (Lindahl 1996; Gilbert et al. 2003a). Region- and position-specific damage rates add to the problem (Gilbert et al. 2003a). Unfortunately, DNA survival and postmortem damage of DNA of a particular sample are unpredictable and difficult to assess. The indirect determination of ancient nucleotide sequences is accompanied by very high standards that must be safeguarded to "guarantee" authenticity of the results (summarized by Cooper and Poinar 2000).
Here, we show that one important aspect is neglected in the discussion on the authenticity of PCR-based results, i.e., the mutagenic effect of ancient DNA extracts themselves. We have noticed such mutagenic effects of ancient DNA extracts in another study (Pusch et al., in preparation). Here, we present evidence that artificial (i.e., nonauthentic) HVRI sequences can be obtained from modern human DNA templates when spiked with ancient DNA extracts. The results indicate that the authenticity of PCR-generated ancient DNA sequences must be reconsidered and illustrate the need for developing standards for an assessment of the mutagenic effect of ancient DNA extracts prior to analysis.
Material and Methods
DNA Extractions
The contemporary human genomic DNA was extracted according to the instructions of the QIAamp DNA Blood Kit (Qiagen, Valencia, Calif.). Ancient DNA extracts were prepared according to a slightly modified "silica-based protocol" (H?ss and P??bo 1993). Bone powder was added to 1 ml extraction buffer (5 M guanidinium thiocyanate, 0.1 M Tris–HCl, 0.02 M EDTA, 1.3% Triton-X-100 pH 7.4) and incubated at 60°C for at least several hours. After centrifugation for 5 min at 5.000 r.p.m., 500 μl of the supernatant was added to 500 μl extraction buffer and 40 μl silica suspension (Boom et al. 1990). Following incubation at room temperature for 10 min, the silica was washed twice with 5 M guanidinium thiocyanate, 0.1 M Tris–HCl, pH 7.4 and twice with 70% ethanol, and the DNA was eluted with water from the air-dried pellet.
For some initial control experiments, the "voltage-induced release method" (Bachmann et al. 2000) and the "mix-and-clean method" (Scholz and Pusch 1997) were used. In short, the voltage-induced release method is based on an electrophoretic separation using 2% w/v SeparideTM (Gibco BRL Life Technologies, Carlsbad, Calif.) gels in 1 X Tris-borate-EDTA (TBE) buffer. A maximum of 25–30 mg ground sample is loaded into a slot along with 25 μl of 1.5 X loading buffer (15% glycerol, 3% SDS, 150 mM dithiothreitol, and 50 mM Tris–HCl, pH 7.5). After gentle stirring, electrophoresis is carried out at 7 V/cm until the xylenecyanol/bromophenol blue dye loaded into empty lanes moved into the gel. The DNA is subsequently isolated from the gel following Pusch's method (1997). For the mix-and-clean method up to 0.2 g, the ground sample is vigorously vortexed for 1 min at 60°C in 500 μl 8% sucrose, 5% Triton-X-100, 10 mM EDTA, and 5 mM Tris–HCl, pH 8.0 buffer. After adding 500 μl phenol, the mixture is incubated for 5 h at room temperature on a shaker. Following an extraction of the aqueous phase with 1 vol chloroform the DNA is precipitated with 0.7 vol propan-2-ol and 20 μg glycogen. The air-dried DNA pellet is resuspended in 8%, 0.1% Triton-X-100, 5 mM EDTA, and 1.2 M NaCl, pH 8.0. The precipitation and resuspension of DNA is repeated twice before the DNA is finally dissolved in TE buffer.
PCR Amplifications for Subsequent Sequencing
Human mitochondrial HVRI (16147–16294) was amplified for 30 cycles. Each cycle consisted of 30-s denaturation at 94°C, 30-s annealing between 55°C and 63°C (5 cycles at 63°C, 5 cycles at 59°C, and 20 cycles at 55°C), and 30-s extension at 72°C. A final extension step of 10 min at 72°C terminated the program. The PCR mix (50 μl) consisted of 0.2 mM of each dNTP, 0.2 μM of each primer, 1 U Ampli-Taq polymerase (Applied Biosystems, Foster City, Calif.), and 50 ng contemporary human DNA in 30 mM MgCl2, 500 mM KCl, and 100 mM Tris, pH 8.9. To test the effect of various ancient DNA extracts on PCR fidelity, 1 μl water was replaced by 1 μl ancient DNA extract. The primers used for amplification were: L16170 5'-CCACCTGTAGTACATAAAAACCCA-3' and H16271 5'-GTGGGTAGGTTTGTTGGTATCCTA-3' for the full-size products, L16170 and H16245 5'-TGAGGGGTGGCTTTGGAGTTG-3' for the short left side (sls) products, and L16182 5'-TACATAAAAACCCAATCCACATCAAA-3' and H16271 for the short right side (srs) products.
PCR Amplifications for Length-Inhibition Analysis
To study the inhibitory effect of ancient DNA extracts on strand elongation during PCR the human mitochondrial region encompassing the ATPase 6, cytochrome oxidase subunit 3, tRNA-Gly, and NADH dehydrogenase subunit 3 (Anderson et al. 1981) was targeted. One reverse primer (H strand) was used in combination with seven forward primers (L strand) in order to amplify fragments of increasing length. The PCR mixture was as the one described earlier. The PCR protocol for cycling consisted of 30 cycles with 30 s denaturation at 94°C, 30 s annealing between 55°C and 60°C (5 cycles at 60°C, 5 cycles at 58°C, and 20 cycles at 55°C), and 30-s extension at 72°C. A final 10-min extension at 72°C terminated the program. The following primers were used for amplification: H10152 5'-CTATGTAGCCGTTGAGTTGTG-3', L9952 5'-GACTATTTCTGTATGTCTCCATC-3' (201 bp), L9747 5'-ACTTCGAGTCTCCCTTCACCA-3' (406 bp), L9501 5'-TGAGCCTTTTACCACTCCAGC-3' (652 bp), L9301 5'-CCATGTGATTTCACTTCCACTC-3' (852 bp), L9081 5'-CCTTCCCTCTACACTTATCATC-3' (1072 bp), L8901 5'-AGCCCACTTCTTACCACAAGG-3' (1252 bp), and L8650 5'-CTAATCACCACCCAACAATGAC-3' (1503 bp).
Plasmid Cloning and DNA Sequencing
The obtained PCR products were cleaned by means of the Concert Gel Extraction System (Invitrogen) and subsequently cloned using a TA cloning kit (Invitrogen). Plasmid DNA from positive colonies was prepared by means of the Plasmid Mini Kit (Qiagen) and subsequently sequenced using BigDye chemistry (Applied Biosystems).
Results
The nucleotide sequence of a 148-bp stretch of the mitochondrial HVRI (16147–16294) of a contemporary human male was sequenced and found to be identical to the Cambridge Reference Sequence (CRS) published by Anderson et al. (1981). The same HVRI stretch was then repeatedly amplified from the same contemporary DNA template but spiked with a series of 14 selected ancient DNA extracts listed in table 1. In most cases, a significant inhibitory effect of the ancient DNA extract on the Taq polymerase activity was observed (table 1).
Table 1 Samples Used for Spiking PCR Amplifications of a 148-bp Stretch of the Mitochondrial HVRI (16147–16294) of a Contemporary Human Template DNA and the Observed Interim Consensus Sequences.
Inhibitory Effect of Ancient DNA Extracts
To quantify the inhibitory effect, 201-, 406-, 652-, 852-p, 1,072-, 1,252-, and 1,503-bp stretches of the human mitochondrial region encompassing the ATPase 6, cytochrome oxidase subunit 3, tRNA-Gly, and NADH dehydrogenase subunit 3 were targeted after spiking the contemporary human template DNA with diluted ancient DNA extracts. It turned out that the inhibitory effect of ancient DNA extracts does not only affect the yield of PCR products but, in some instances, also the length of the fragment that can be amplified. For example, figure 1 shows a dilution of 1:100 of the extract from an approximately 12,000-year-old tooth of Equus spp is necessary in order to successfully amplify the 201-bp fragment from a 50-ng modern human template DNA. Spiking with a 1:1,000,000 dilution of the Equus spp extract does still not allow the successful amplification of an 852-bp fragment. All fragments amplified equally from the modern DNA template in the control experiments without spiking (fig. 1).
FIG. 1. Amplification of seven PCR products (201, 406, 652, 852, 1,072, 1,252, and 1,503 bp) of the human mitochondrial region encompassing the ATPase 6, cytochrome oxidase subunit 3, tRNA-Gly and NADH dehydrogenase subunit 3 using modern human DNA that was spiked with various dilutions of a DNA extract from an approximately 12,000-year-old tooth of Equus spp
Mutagenic Effect of Ancient DNA Extracts
After determining the required dilutions of the 14 ancient DNA extracts used for the spiking experiments (table 1), the obtained PCR products of the targeted 148-bp stretch of the mitochondrial HVRI (16147–16294) were cloned and sequenced. A total of 547 HVRI sequences were subsequently analyzed. Only 23.03% (126 sequences) were identical to CRS and were, therefore, correct. The vast majority of sequences (76.97%) differed from CRS by single nucleotide substitutions and/or insertions/deletions (two examples are shown in fig. 2). Some of these mutations were clone specific, whereas others were shared by several or all clones of a particular amplicon, indicating that mutations were introduced at different stages of the PCR. After eliminating all clone-specific alterations, leaving only mutations that were shared by at least two clones of a particular amplicon, the interim consensus sequences were collected for each amplicon. It was not our intention to derive the hypothetical authentic sequences of the amplicons, but to comprehensively assess the observed sequence diversity. The deduced 35 interim consensus sequences were subsequently aligned to CRS (table 2). To improve the alignment, three gaps had to be assumed, extending the alignment to 103 positions (PCR primers excluded). The interim consensus sequences that are most different from CRS (XXVII and XXVIII) differ from CRS by as many as 18 positions. Interim consensus sequence XXVI that represents 86 sequences (15.72%) still differs from CRS at 16 positions.
FIG. 2. Examples of altered human sequences of cloned PCR products of a 148-bp stretch of mitochondrial HVRI (16147–16294) generated in spiking experiments with ancient DNA extracts. The upper sequence is altered at 6 positions and the lower one at 18 positions. The correct nucleotides of the human PCR template (identical to CRS) are indicated at the respective positions under the sequence
Table 2 Observed Interim Consensus Sequences of Cloned PCR Products of a 148-bp Stretch of Mitochondrial HVRI (16147–16294) of a Contemporary Human Amplified Under the Presence of Various Ancient DNA extracts (table 1).
The observed sequence alterations are not randomly distributed, i.e., interim consensus sequences of amplicons generated through independent amplifications spiked with different ancient DNA extracts share similar nucleotide substitutions. In total, 34 positions are affected by 38 mutations of which 25 are transitions, 6 are transversions, and 7 are insertions/deletions. Twenty of these mutations occur in less than 10%, 7 in 10%–30%, and 11 in more than 30% of all clones analyzed (table 3).
Table 3 Frequency of Mutations Listed in Table 2 Among Cloned PCR Products of a 148-bp Stretch of Mitochondrial HVRI (16147–16294) of Spiked Contemporary Human DNA.
Many of the observed mutations affect positions that are variable in human populations and were previously detected in either contemporary or prehistoric human samples. The CT transition at position 16223 (19 interim consensus sequences in table 2) that was detected in the Pagglici 12 remains of a 24,000-year-old anatomically modern human from the Pagglici cave in southern Italy (Caramelli et al. 2003) and the TC transition at position 16224 (interim consensus sequences II and XIV in table 2) that was detected in the Tyrolean iceman (Handt et al. 1994) are two examples. Interim consensus sequence XXVI (table 2), which represents 15.7% of all analyzed sequences, differs from the closest human match in GenBank by nine substitutions and three gaps but shares 7 out of 11 mutations that are in combination characteristic for the Neandertal sequence AF011222 (Krings et al. 1997). All 7 mutations belong to those that occur in more than 30% of the analyzed clones. Furthermore, eight positions were previously identified as mutation hotspots that are likely to be affected by postmortem damage (Gilbert et al. 2003b).
To test if the observed sequence alterations were caused by the DNA molecules in the ancient DNA extracts, six ancient DNA extracts (i.e., Falco spp, rodent coprolite, Sequoia spp, and Homo erectus listed in table 1, and Araucaria spp (wood) from Madagascar (190–230 Myr), and Carcharocles angustidens (tooth) from the USA (23–28 Myr)) were pretreated with DNaseI prior to the spiking experiments. A total of 67 sequences were deduced and the majority of them (71.6%) were altered. Many mutations were identical to those observed previously and were grouped into 14 different interim consensus sequences (fig. 3).
FIG. 3. Observed consensus sequences of cloned PCR products of a 148-bp stretch of mitochondrial HVRI (16147–16294) of a contemporary human amplified under the presence of six ancient DNA extracts digested with DNaseI prior to amplification. Dashes indicate gaps introduced in order to improve the alignment
Discussion
We present experimental evidence that ancient DNA extracts may have a strong inhibitory effect on PCR amplifications (table 1). In spiking experiments with ancient DNA extracts of various age and origin, 3 out of 14 extracts required a 1:10,000 dilution in order to restore full Taq polymerase activity for a targeted 148-bp amplicon. In our collection of ancient DNA extracts, the extent of the inhibitory effect varied considerably, e.g., five samples showed no inhibitory effect at all. The inhibitory potential of ancient DNA extracts on PCR amplifications was unpredictable and could not be related to the age of a sample, i.e., the inhibitory effect had to be determined experimentally for each particular extract.
Our experiments show that failure of amplification of large fragments does not necessarily indicate absence of high molecular weight template DNA in the PCR mixture (in our inhibition studies by spiking experiments all PCR mixtures contained high molecular weight human template DNA). This argument is frequently applied in the literature (e.g., Caramelli et al. 2003, Poinar et al. 2003) in order to support lack of contamination in ancient DNA extracts according to the "appropriate molecular behavior" criterion of Cooper and Poinar (2000), i.e., large 500–1,000 base pair products are unusual.
We also present experimental evidence that ancient DNA extracts may induce mutations under PCR amplifications. The strength of the "spiking contemporary human DNA templates with ancient DNA extracts approach" is that properties of ancient DNA extracts can be studied in a modern DNA experimental setup. Doing so, the shortcomings of ancient DNA experiments can be avoided. The nucleotide sequence of the modern PCR template used in our experiments can be determined unambiguously, i.e., it is identical to the human mtDNA determined by Anderson et al. (1981) when amplified without spiking. The approach allows us to conclude that the frequently observed sequence alterations in sequences from spiked amplifications must have occurred under PCR.
Analysis of 547 sequences from cloned amplicons of a 148-bp stretch (16147–16294) of the mitochondrial HVRI indicates that extract-induced mutations do not occur at random. In total, 34 positions of a 103-bp alignment (PCR primers excluded) are affected, the majority of those repeatedly in independent PCR amplifications. However, this does not mean that extract-induced mutations can predictably be reproduced. It is more that particular positions of the target sequence have a high likelihood of being altered.
It is certainly difficult to identify the cause of the observed sequence heterogeneity. Intuitively, one might suspect human DNA in the ancient DNA extracts used for spiking (note: in this context it is irrelevant whether such human DNA is authentic for a particular find or results from contamination). Such DNA might serve as a proper template and differ in sequence from targeted human DNA of a known sequence. Alternatively, such DNA might be degenerated and, therefore, not serve as proper template but cause sequence variation in the obtained amplicons through the process of PCR jumping (P??bo 1989). We believe that this line of argument has to be rejected for the following reasons. First, the majority of DNA extracts used for spiking had to be diluted prior to PCR amplification (table 1). Only 1 μl diluted extract was then added to 50 ng modern human template DNA; i.e., the number of modern human template DNA molecules in the PCR mixture therefore exceeded by far putative human DNA molecules from the spiking extracts by several orders of magnitudes. Second, human DNA from other sources than the targeted modern human DNA can hardly explain why 11 mutations were found in more than 30% of the analyzed clones (table 3). Third, human DNA from other sources than the targeted modern human DNA can hardly explain that a combination of, e.g., 16 mutations (interim consensus sequence XXVI in table 2) occurs in 86 clones, i.e., 15.7% of all sequences, generated in seven different spiking experiments. Fourth, human DNA from other sources than the targeted modern human DNA can hardly explain why many interim consensus sequences occur that do not match any HVRI sequences deposited in sequence databases. Finally, spiking experiments using six ancient DNA extracts pretreated with DNaseI (67 sequences) also yielded mutated sequences with similar nucleotide alterations. Accordingly, DNA could be excluded as the mutagenic agent.
We speculate that coextracted chemicals other than DNA might induce the observed sequence alterations. Such coextracted mutagenic substances may accumulate in the samples during diagenesis. Although diagenetic processes are barely understood, we know that specimens accumulate postmortem multivalent metal ions. For example, concentrations of manganese, barium, strontium, iron, and uranium can be enriched up to 5,000 times compared with contemporary samples (Bischoff 1981, Velasco-Vazquez et al. 1997, Kohn et al. 1999). In particular, manganese is known for its mutagenic effect, and molecular biologists use this effect when using mutagenesis kits. In preliminary studies, we noticed substantially elevated manganese concentrations for the samples used here (data not shown). In contrast, manganese was not detected in 87 modern controls. If coextracted manganese is the reason for the observed mutations, one would, at least at first glance, expect to see a predictive relationship between manganese concentration and reliability of sequences. However, there is no such relationship, and, even worse, there is no reason to expect such a simple correlation. This is because the chemistry of manganese is very complex. The element occurs with a variety of valence stages that differ in their biological activity. However, manganese measurements refer to the total amount of manganese and cannot differentiate valence stages. In addition, the mutagenic effect of manganese also depends on the concentration of other ions. Svetlov and Cooper (1998) and Kunichika et al. (2002), for example, noticed that the mutagenic effect of manganese changed with changing concentration of magnesium. Other reasons may be added (see, e.g., Fletcher et al. 1994, Pelletier et al. 1996). Nevertheless, Pusch et al. (in preparation) showed in other experiments that a GA transition occurs frequently at cDNA position 1,138 of the human FGFR3 gene in PCR amplifications under the presence of 0.25 mM MnCl2. The same diagnostic mutation also occurs when contemporary human template DNA is spiked with an ancient DNA extract from a Pleistocene Ursus spelaeus bone. Future research will have to address this topic.
If coextracted chemicals such as multivalent metal ions were the reason for the observed sequence alterations, they might be removable by extraction methods, which are based on different extraction principles than the default silica-based protocol. In preliminary experiments, we also tested two other extraction methods, i.e., the "voltage-induced release method" (Bachmann et al. 2000) and the "mix-and-clean method" (Scholz and Pusch 1997). However, DNA extracted according to these alternative methods also yielded altered sequences in spiking experiments (data not shown). Future studies will have to address why hypothetical coextracted mutagenic substances cannot be removed by currently applied methods.
The data presented here have implications for ancient DNA research and may challenge the authenticity of a number of PCR-generated sequences from ancient samples. For some reasons it is, however, impossible to assess how frequently sequence alterations induced by ancient DNA extracts may have affected published sequences. Many papers report only the deduced consensus sequences and neither the sequences of the individual clones nor information on the observed sequence variation. Furthermore, as can be seen from table 1, different ancient DNA extracts induce sequence alterations to a different extent. For some extracts, the majority of sequences were identical to CRS and the obtained mutations occurred at very low frequencies. In those cases, extract-induced sequence alterations would not make it into a consensus sequence when applying the majority rule (hypothetical consensus sequence A in tables 1 and 4). In other instances, the majority of sequences were altered and the deduced consensus sequences wrong (hypothetical consensus sequences B–J in tables 1 and 4). It might also be a matter of chance which mutations are considered to be authentic, because usually the clones used for analysis are selected at random.
Table 4 Deduced Hypothetical Consensus Sequences of Cloned PCR Products of a 148-bp Stretch of Mitochondrial HVRI (16147–16294) of a Contemporary Human Amplified Under the Presence of Various Ancient DNA Extracts (table 1) (CRS is the correct sequence).
Intuitively, one might feel that applying the rule to clone and sequence short overlapping PCR products from several independent amplifications in order to deduce a consensus sequence may prevent from pitfalls. According to our data, this assumption is not justified. In the data set presented here 11 alterations occurred in more than 30% of the analyzed clones and some of those would make it into a consensus sequence in an ancient DNA approach when following the guidelines for ancient DNA. Independent replication of the experiments in other laboratories to verify deduced consensus sequences of ancient samples are of no help since such mutations can, as shown, repeatedly occur in independent amplifications. As stated earlier, particular positions have a high likelihood of being altered.
Our data imply that an unknown number of published ancient DNA sequences have to be considered with caution. There is an urgent need to identify the causes for the observed mutations in the spiking experiments and to develop standards for an assessment of the mutagenic effect of ancient DNA extracts as a mandatory criterion for authenticity. Although to a different extent in different studies, it is likely that extract-induced mutations are an intrinsic and a general problem of PCR amplifications of ancient templates.
Acknowledgements
We are grateful to M. Broghammer for excellent technical assistance. This work was supported in part by grants from the Research Council of Norway (National Centre for Biosystematics, 146515/420), the German Federal Ministry of Education, Science, Research and Technology (F?. 01KS9602), and the Interdisciplinary Center of Clinical Research (IZKF-Q3) Tübingen.
Literature Cited
Anderson, S., A. T. Bankier, and B. G. Barrel, et al. (14 coauthors). 1981. Sequence and organisation of the human mitochondrial genome. Nature 290:457-465.
Bachmann, L., M. Scholz, M. Broghammer, I. Giddings, and C. M. Pusch. 2000. Voltage-induced release of nucleic acids from palaeontological samples. Electrophoresis 21:1488-1492.
Bischoff, J. L. 1981. A geochemical study of the Eyasi I skull: implications for the process of fossilization. Pp. 128–132 in: H. Müller-Beck, ed. The palaeoanthropological finds of the Pliocene and Pleistocene. Archaeologica Venatoria, Tübingen-Bad Bramstedt.
Boom, R., C. J. Sol, M. M. Salimans, C. L. Jansen, P. M. Wertheim-van Dillen, and J. van der Noordaa. 1990. Rapid and simple method for purification of nucleic acids. J. Clin. Microbiol. 28:495-503.
Caramelli, D., C. Lalueza-Fox, and C. Vernesi, et al. (11 coauthors). 2003. Evidence for a genetic discontinuity between Neandertals and 24,000-year-old anatomically modern Europeans. Proc. Natl. Acad. Sci. USA. 27:6593-6597.
Cooper, A., and H. N. Poinar. 2000. Ancient DNA: do it right or not at all. Science 289:1139.
Fletcher, G. G., F. E. Rossetto, J. D. Turnbull, and E. Nieboer. 1994. Toxicity, uptake, and mutagenicity of particulate and soluble nickel compounds. Environ. Health Perspect. 102:(Suppl.) 3: 69-79.
Gilbert, M.T., A. J. Hansen, E. Willerslev, L. Rudbeck, I. Barnes, N. Lynnerup, and A. Cooper. 2003a. Characterization of genetic miscoding lesions caused by postmortem damage. Am. J. Hum. Genet. 72:48-61.
Gilbert, M. T., E. Willerslev, A. J. Hansen, I. Barnes, L. Rudbeck, N. Lynnerup, and A. Cooper. 2003b. Distribution patterns of postmortem damage in human mitochondrial DNA. Am. J. Hum. Genet. 72:32-47 Erratum in: Am. J. Hum. Genet. 72:779.
Handt O., M. Richards, and M. Trommsdorff, et al. (13 coauthors). 1994. Molecular genetic analyses of the Tyrolean Ice Man. Science 264:1775-1778.
Hansen, A. J., E. Willerslev, C. Wiuf, T. Mourier, and P. Arctander. 2001. Statistical evidence for miscoding lesions in ancient DNA templates. Mol. Biol. Evol. 18:262-265.
Hofreiter, M., V. J?nicke, D. Serre, A. von Haeseler, and S. P??bo. 2001a. DNA sequences from multiple amplifications reveal artifacts induced by cytosine deamination in ancient DNA. Nucleic Acids Res. 29:4793-4799.
Hofreiter, M., D. Serre, H. N. Poinar, M. Kuch, and S. P??bo. 2001b. Ancient DNA. Nature Rev. Genet. 2:353-359.
H?ss, M., and S. P??bo. 1993. DNA extraction from Pleistocene bones by a silica-based purification method. Nucleic Acids Res. 21:3913-3914.
Kohn, M. J., M. J. Schoeninger, and W. W. Barker. 1999. Altered states: effects of diagenesis on fossil tooth chemistry. Geochim. Cosmochim. Acta 63:2737-2747.
Krings, M., A. Stone, R. W. Schmitz, H. Krainitzki, M. Stoneking, and S. P??bo. 1997. Neandertal DNA sequences and the origin of modern humans. Cell 90:19-30.
Kunichika, K., H. Hashimoto, and T. Imoto. 2002. Robustness of hen lysozyme monitored by random mutations. Protein Eng. 15:805-809.
Lindahl, T. 1996. The Croonian Lecture, 1996: endogenous damage to DNA. Philos. Trans. R. Soc. Lond. B Biol. Sci. 351:1529-1538.
P??bo, S. 1989. Ancient DNA: extraction, characterization, molecular cloning, and enzymatic amplification. Proc. Natl. Acad. Sci. USA 86:1939-1943.
Pelletier, H., M. R. Sawaya, W. Wolfle, S. H. Wilson, and J. Kraut. 1996. A structural basis for metal ion mutagenicity and nucleotide selectivity in human DNA polymerase beta. Biochemistry 35:12762-12777.
Poinar, H., M. Kuch, G. McDonald, P. Martin, and S. P??bo. 2003. Nuclear gene sequences from a late pleistocene sloth coprolite. Curr. Biol. 13:1150-1152.
Pusch, C. M. 1997. A simple and fast procedure for high quality DNA isolation from gels using laundry detergent and inverted columns. Electrophoresis 18:1103-1104.
Pusch, C. M., L. Bachmann, M. Broghammer, and M. Scholz. 2000. Internal Alu-polymerase chain reaction: a sensitive contamination monitoring protocol for DNA extracted from prehistoric animal bones. Anal. Biochem. 284:408-411.
Scholz, M., and C. Pusch. 1997. An efficient isolation method for high-quality DNA from ancient bones. Trends Genet. 13:249.
Svetlov, V., and T. G. Cooper. 1998. Efficient PCR-based random mutagenesis of sub-genic (100bp) DNA fragments. Yeast 14:89-91.
Velasco-Vazquez, J., M. Arnay-de-la-Rosa, E. Gonzalez-Reimers, and O. Hernandez-Torres. 1997. Paleodietary analysis on the prehistoric population of El Hierro (Canary Islands). Biol. Trace Elem. Res. 60:235-241.
Willerslev, E., A. J. Hansen, B. Christensen, J. P. Steffensen, and P. Arctander. 1999. Diversity of Holocene life forms in fossil glacier ice. Proc. Natl. Acad. Sci. USA 96:8017-8021.(Carsten M. Pusch* and Lut)
Department of Zoology, Natural History Museums and Botanical Garden, University of Oslo, Oslo, Norway
E-mail: bachmann@nhm.uio.no.
Abstract
Proof of authenticity is the greatest challenge in palaeogenetic research, and many safeguards have become standard routine in laboratories specialized on ancient DNA research. Here we describe an as-yet unknown source of artifacts that will require special attention in the future. We show that ancient DNA extracts on their own can have an inhibitory and mutagenic effect under PCR. We have spiked PCR reactions including known human test DNA with 14 selected ancient DNA extracts from human and nonhuman sources. We find that the ancient DNA extracts inhibit the amplification of large fragments to different degrees, suggesting that the usual control against contaminations, i.e., the absence of long amplifiable fragments, is not sufficient. But even more important, we find that the extracts induce mutations in a nonrandom fashion. We have amplified a 148-bp stretch of the mitochondrial HVRI from contemporary human template DNA in spiked PCR reactions. Subsequent analysis of 547 sequences from cloned amplicons revealed that the vast majority (76.97%) differed from the correct sequence by single nucleotide substitutions and/or indels. In total, 34 positions of a 103-bp alignment are affected, and most mutations occur repeatedly in independent PCR amplifications. Several of the induced mutations occur at positions that have previously been detected in studies of ancient hominid sequences, including the Neandertal sequences. Our data imply that PCR-induced mutations are likely to be an intrinsic and general problem of PCR amplifications of ancient templates. Therefore, ancient DNA sequences should be considered with caution, at least as long as the molecular basis for the extract-induced mutations is not understood.
Key Words: ancient DNA extracts ? mitochondrial DNA ? mutation ? PCR errors ? PCR inhibition
Introduction
The retrieval of nucleotide sequences from ancient DNA templates requires technical skills and extensive expertise. The major shortcoming of ancient DNA research is that authentic sequences can only be determined in an indirect way. This is particularly true for samples from extinct taxa, i.e., when no high-quality DNA sample is available. To overcome the problem, short overlapping PCR products are cloned, and the authentic sequence is subsequently deduced as the consensus of many individual sequences that are derived from several independent amplifications (Cooper and Poinar, 2000; Hofreiter et al. 2001a,b). Such cloned PCR products are usually heterogeneous due to contamination with contemporary DNA, PCR artifacts, and assumed postmortem damage of authentic DNA molecules.
Contamination with modern DNA has plagued ancient DNA research, and this is a matter of particular concern when working with hominid samples. It might be impossible to prove authenticity of the obtained data simply because of the expected very high level of sequence similarity between authentic ancient DNA from a hominid fossil and contaminating contemporary DNA. Even minute amounts of contaminating DNA might question the results obtained by amplification of ancient DNA using the extremely sensitive polymerase chain reaction (PCR). However, it must be stressed that contamination is a general problem in ancient DNA research and affects the analysis of samples from any taxon (Pusch et al. 2000). Usually, PCR protocols are optimized using DNA from closely related taxa and carryover contaminations might occur.
PCR artifacts such as PCR jumping, allelic dropouts, regular polymerase errors, and assumed postmortem damage of authentic DNA molecules are also great obstacles in the retrieval of authentic sequence data from ancient DNA (P??bo 1989; Willerslev et al. 1999; Hansen et al. 2001; Gilbert et al. 2003a). Strand breaks decrease the average molecular weight of nucleic acids (P??bo 1989). Oxidative and hydrolytic damage may lead to incorporation of incorrect nucleotides or block strand elongation during PCR (Willerslev et al. 1999). The conversion of adenine to hypoxanthine and cytosine to uracil, respectively, are considered common forms of DNA damage (Lindahl 1996; Gilbert et al. 2003a). Region- and position-specific damage rates add to the problem (Gilbert et al. 2003a). Unfortunately, DNA survival and postmortem damage of DNA of a particular sample are unpredictable and difficult to assess. The indirect determination of ancient nucleotide sequences is accompanied by very high standards that must be safeguarded to "guarantee" authenticity of the results (summarized by Cooper and Poinar 2000).
Here, we show that one important aspect is neglected in the discussion on the authenticity of PCR-based results, i.e., the mutagenic effect of ancient DNA extracts themselves. We have noticed such mutagenic effects of ancient DNA extracts in another study (Pusch et al., in preparation). Here, we present evidence that artificial (i.e., nonauthentic) HVRI sequences can be obtained from modern human DNA templates when spiked with ancient DNA extracts. The results indicate that the authenticity of PCR-generated ancient DNA sequences must be reconsidered and illustrate the need for developing standards for an assessment of the mutagenic effect of ancient DNA extracts prior to analysis.
Material and Methods
DNA Extractions
The contemporary human genomic DNA was extracted according to the instructions of the QIAamp DNA Blood Kit (Qiagen, Valencia, Calif.). Ancient DNA extracts were prepared according to a slightly modified "silica-based protocol" (H?ss and P??bo 1993). Bone powder was added to 1 ml extraction buffer (5 M guanidinium thiocyanate, 0.1 M Tris–HCl, 0.02 M EDTA, 1.3% Triton-X-100 pH 7.4) and incubated at 60°C for at least several hours. After centrifugation for 5 min at 5.000 r.p.m., 500 μl of the supernatant was added to 500 μl extraction buffer and 40 μl silica suspension (Boom et al. 1990). Following incubation at room temperature for 10 min, the silica was washed twice with 5 M guanidinium thiocyanate, 0.1 M Tris–HCl, pH 7.4 and twice with 70% ethanol, and the DNA was eluted with water from the air-dried pellet.
For some initial control experiments, the "voltage-induced release method" (Bachmann et al. 2000) and the "mix-and-clean method" (Scholz and Pusch 1997) were used. In short, the voltage-induced release method is based on an electrophoretic separation using 2% w/v SeparideTM (Gibco BRL Life Technologies, Carlsbad, Calif.) gels in 1 X Tris-borate-EDTA (TBE) buffer. A maximum of 25–30 mg ground sample is loaded into a slot along with 25 μl of 1.5 X loading buffer (15% glycerol, 3% SDS, 150 mM dithiothreitol, and 50 mM Tris–HCl, pH 7.5). After gentle stirring, electrophoresis is carried out at 7 V/cm until the xylenecyanol/bromophenol blue dye loaded into empty lanes moved into the gel. The DNA is subsequently isolated from the gel following Pusch's method (1997). For the mix-and-clean method up to 0.2 g, the ground sample is vigorously vortexed for 1 min at 60°C in 500 μl 8% sucrose, 5% Triton-X-100, 10 mM EDTA, and 5 mM Tris–HCl, pH 8.0 buffer. After adding 500 μl phenol, the mixture is incubated for 5 h at room temperature on a shaker. Following an extraction of the aqueous phase with 1 vol chloroform the DNA is precipitated with 0.7 vol propan-2-ol and 20 μg glycogen. The air-dried DNA pellet is resuspended in 8%, 0.1% Triton-X-100, 5 mM EDTA, and 1.2 M NaCl, pH 8.0. The precipitation and resuspension of DNA is repeated twice before the DNA is finally dissolved in TE buffer.
PCR Amplifications for Subsequent Sequencing
Human mitochondrial HVRI (16147–16294) was amplified for 30 cycles. Each cycle consisted of 30-s denaturation at 94°C, 30-s annealing between 55°C and 63°C (5 cycles at 63°C, 5 cycles at 59°C, and 20 cycles at 55°C), and 30-s extension at 72°C. A final extension step of 10 min at 72°C terminated the program. The PCR mix (50 μl) consisted of 0.2 mM of each dNTP, 0.2 μM of each primer, 1 U Ampli-Taq polymerase (Applied Biosystems, Foster City, Calif.), and 50 ng contemporary human DNA in 30 mM MgCl2, 500 mM KCl, and 100 mM Tris, pH 8.9. To test the effect of various ancient DNA extracts on PCR fidelity, 1 μl water was replaced by 1 μl ancient DNA extract. The primers used for amplification were: L16170 5'-CCACCTGTAGTACATAAAAACCCA-3' and H16271 5'-GTGGGTAGGTTTGTTGGTATCCTA-3' for the full-size products, L16170 and H16245 5'-TGAGGGGTGGCTTTGGAGTTG-3' for the short left side (sls) products, and L16182 5'-TACATAAAAACCCAATCCACATCAAA-3' and H16271 for the short right side (srs) products.
PCR Amplifications for Length-Inhibition Analysis
To study the inhibitory effect of ancient DNA extracts on strand elongation during PCR the human mitochondrial region encompassing the ATPase 6, cytochrome oxidase subunit 3, tRNA-Gly, and NADH dehydrogenase subunit 3 (Anderson et al. 1981) was targeted. One reverse primer (H strand) was used in combination with seven forward primers (L strand) in order to amplify fragments of increasing length. The PCR mixture was as the one described earlier. The PCR protocol for cycling consisted of 30 cycles with 30 s denaturation at 94°C, 30 s annealing between 55°C and 60°C (5 cycles at 60°C, 5 cycles at 58°C, and 20 cycles at 55°C), and 30-s extension at 72°C. A final 10-min extension at 72°C terminated the program. The following primers were used for amplification: H10152 5'-CTATGTAGCCGTTGAGTTGTG-3', L9952 5'-GACTATTTCTGTATGTCTCCATC-3' (201 bp), L9747 5'-ACTTCGAGTCTCCCTTCACCA-3' (406 bp), L9501 5'-TGAGCCTTTTACCACTCCAGC-3' (652 bp), L9301 5'-CCATGTGATTTCACTTCCACTC-3' (852 bp), L9081 5'-CCTTCCCTCTACACTTATCATC-3' (1072 bp), L8901 5'-AGCCCACTTCTTACCACAAGG-3' (1252 bp), and L8650 5'-CTAATCACCACCCAACAATGAC-3' (1503 bp).
Plasmid Cloning and DNA Sequencing
The obtained PCR products were cleaned by means of the Concert Gel Extraction System (Invitrogen) and subsequently cloned using a TA cloning kit (Invitrogen). Plasmid DNA from positive colonies was prepared by means of the Plasmid Mini Kit (Qiagen) and subsequently sequenced using BigDye chemistry (Applied Biosystems).
Results
The nucleotide sequence of a 148-bp stretch of the mitochondrial HVRI (16147–16294) of a contemporary human male was sequenced and found to be identical to the Cambridge Reference Sequence (CRS) published by Anderson et al. (1981). The same HVRI stretch was then repeatedly amplified from the same contemporary DNA template but spiked with a series of 14 selected ancient DNA extracts listed in table 1. In most cases, a significant inhibitory effect of the ancient DNA extract on the Taq polymerase activity was observed (table 1).
Table 1 Samples Used for Spiking PCR Amplifications of a 148-bp Stretch of the Mitochondrial HVRI (16147–16294) of a Contemporary Human Template DNA and the Observed Interim Consensus Sequences.
Inhibitory Effect of Ancient DNA Extracts
To quantify the inhibitory effect, 201-, 406-, 652-, 852-p, 1,072-, 1,252-, and 1,503-bp stretches of the human mitochondrial region encompassing the ATPase 6, cytochrome oxidase subunit 3, tRNA-Gly, and NADH dehydrogenase subunit 3 were targeted after spiking the contemporary human template DNA with diluted ancient DNA extracts. It turned out that the inhibitory effect of ancient DNA extracts does not only affect the yield of PCR products but, in some instances, also the length of the fragment that can be amplified. For example, figure 1 shows a dilution of 1:100 of the extract from an approximately 12,000-year-old tooth of Equus spp is necessary in order to successfully amplify the 201-bp fragment from a 50-ng modern human template DNA. Spiking with a 1:1,000,000 dilution of the Equus spp extract does still not allow the successful amplification of an 852-bp fragment. All fragments amplified equally from the modern DNA template in the control experiments without spiking (fig. 1).
FIG. 1. Amplification of seven PCR products (201, 406, 652, 852, 1,072, 1,252, and 1,503 bp) of the human mitochondrial region encompassing the ATPase 6, cytochrome oxidase subunit 3, tRNA-Gly and NADH dehydrogenase subunit 3 using modern human DNA that was spiked with various dilutions of a DNA extract from an approximately 12,000-year-old tooth of Equus spp
Mutagenic Effect of Ancient DNA Extracts
After determining the required dilutions of the 14 ancient DNA extracts used for the spiking experiments (table 1), the obtained PCR products of the targeted 148-bp stretch of the mitochondrial HVRI (16147–16294) were cloned and sequenced. A total of 547 HVRI sequences were subsequently analyzed. Only 23.03% (126 sequences) were identical to CRS and were, therefore, correct. The vast majority of sequences (76.97%) differed from CRS by single nucleotide substitutions and/or insertions/deletions (two examples are shown in fig. 2). Some of these mutations were clone specific, whereas others were shared by several or all clones of a particular amplicon, indicating that mutations were introduced at different stages of the PCR. After eliminating all clone-specific alterations, leaving only mutations that were shared by at least two clones of a particular amplicon, the interim consensus sequences were collected for each amplicon. It was not our intention to derive the hypothetical authentic sequences of the amplicons, but to comprehensively assess the observed sequence diversity. The deduced 35 interim consensus sequences were subsequently aligned to CRS (table 2). To improve the alignment, three gaps had to be assumed, extending the alignment to 103 positions (PCR primers excluded). The interim consensus sequences that are most different from CRS (XXVII and XXVIII) differ from CRS by as many as 18 positions. Interim consensus sequence XXVI that represents 86 sequences (15.72%) still differs from CRS at 16 positions.
FIG. 2. Examples of altered human sequences of cloned PCR products of a 148-bp stretch of mitochondrial HVRI (16147–16294) generated in spiking experiments with ancient DNA extracts. The upper sequence is altered at 6 positions and the lower one at 18 positions. The correct nucleotides of the human PCR template (identical to CRS) are indicated at the respective positions under the sequence
Table 2 Observed Interim Consensus Sequences of Cloned PCR Products of a 148-bp Stretch of Mitochondrial HVRI (16147–16294) of a Contemporary Human Amplified Under the Presence of Various Ancient DNA extracts (table 1).
The observed sequence alterations are not randomly distributed, i.e., interim consensus sequences of amplicons generated through independent amplifications spiked with different ancient DNA extracts share similar nucleotide substitutions. In total, 34 positions are affected by 38 mutations of which 25 are transitions, 6 are transversions, and 7 are insertions/deletions. Twenty of these mutations occur in less than 10%, 7 in 10%–30%, and 11 in more than 30% of all clones analyzed (table 3).
Table 3 Frequency of Mutations Listed in Table 2 Among Cloned PCR Products of a 148-bp Stretch of Mitochondrial HVRI (16147–16294) of Spiked Contemporary Human DNA.
Many of the observed mutations affect positions that are variable in human populations and were previously detected in either contemporary or prehistoric human samples. The CT transition at position 16223 (19 interim consensus sequences in table 2) that was detected in the Pagglici 12 remains of a 24,000-year-old anatomically modern human from the Pagglici cave in southern Italy (Caramelli et al. 2003) and the TC transition at position 16224 (interim consensus sequences II and XIV in table 2) that was detected in the Tyrolean iceman (Handt et al. 1994) are two examples. Interim consensus sequence XXVI (table 2), which represents 15.7% of all analyzed sequences, differs from the closest human match in GenBank by nine substitutions and three gaps but shares 7 out of 11 mutations that are in combination characteristic for the Neandertal sequence AF011222 (Krings et al. 1997). All 7 mutations belong to those that occur in more than 30% of the analyzed clones. Furthermore, eight positions were previously identified as mutation hotspots that are likely to be affected by postmortem damage (Gilbert et al. 2003b).
To test if the observed sequence alterations were caused by the DNA molecules in the ancient DNA extracts, six ancient DNA extracts (i.e., Falco spp, rodent coprolite, Sequoia spp, and Homo erectus listed in table 1, and Araucaria spp (wood) from Madagascar (190–230 Myr), and Carcharocles angustidens (tooth) from the USA (23–28 Myr)) were pretreated with DNaseI prior to the spiking experiments. A total of 67 sequences were deduced and the majority of them (71.6%) were altered. Many mutations were identical to those observed previously and were grouped into 14 different interim consensus sequences (fig. 3).
FIG. 3. Observed consensus sequences of cloned PCR products of a 148-bp stretch of mitochondrial HVRI (16147–16294) of a contemporary human amplified under the presence of six ancient DNA extracts digested with DNaseI prior to amplification. Dashes indicate gaps introduced in order to improve the alignment
Discussion
We present experimental evidence that ancient DNA extracts may have a strong inhibitory effect on PCR amplifications (table 1). In spiking experiments with ancient DNA extracts of various age and origin, 3 out of 14 extracts required a 1:10,000 dilution in order to restore full Taq polymerase activity for a targeted 148-bp amplicon. In our collection of ancient DNA extracts, the extent of the inhibitory effect varied considerably, e.g., five samples showed no inhibitory effect at all. The inhibitory potential of ancient DNA extracts on PCR amplifications was unpredictable and could not be related to the age of a sample, i.e., the inhibitory effect had to be determined experimentally for each particular extract.
Our experiments show that failure of amplification of large fragments does not necessarily indicate absence of high molecular weight template DNA in the PCR mixture (in our inhibition studies by spiking experiments all PCR mixtures contained high molecular weight human template DNA). This argument is frequently applied in the literature (e.g., Caramelli et al. 2003, Poinar et al. 2003) in order to support lack of contamination in ancient DNA extracts according to the "appropriate molecular behavior" criterion of Cooper and Poinar (2000), i.e., large 500–1,000 base pair products are unusual.
We also present experimental evidence that ancient DNA extracts may induce mutations under PCR amplifications. The strength of the "spiking contemporary human DNA templates with ancient DNA extracts approach" is that properties of ancient DNA extracts can be studied in a modern DNA experimental setup. Doing so, the shortcomings of ancient DNA experiments can be avoided. The nucleotide sequence of the modern PCR template used in our experiments can be determined unambiguously, i.e., it is identical to the human mtDNA determined by Anderson et al. (1981) when amplified without spiking. The approach allows us to conclude that the frequently observed sequence alterations in sequences from spiked amplifications must have occurred under PCR.
Analysis of 547 sequences from cloned amplicons of a 148-bp stretch (16147–16294) of the mitochondrial HVRI indicates that extract-induced mutations do not occur at random. In total, 34 positions of a 103-bp alignment (PCR primers excluded) are affected, the majority of those repeatedly in independent PCR amplifications. However, this does not mean that extract-induced mutations can predictably be reproduced. It is more that particular positions of the target sequence have a high likelihood of being altered.
It is certainly difficult to identify the cause of the observed sequence heterogeneity. Intuitively, one might suspect human DNA in the ancient DNA extracts used for spiking (note: in this context it is irrelevant whether such human DNA is authentic for a particular find or results from contamination). Such DNA might serve as a proper template and differ in sequence from targeted human DNA of a known sequence. Alternatively, such DNA might be degenerated and, therefore, not serve as proper template but cause sequence variation in the obtained amplicons through the process of PCR jumping (P??bo 1989). We believe that this line of argument has to be rejected for the following reasons. First, the majority of DNA extracts used for spiking had to be diluted prior to PCR amplification (table 1). Only 1 μl diluted extract was then added to 50 ng modern human template DNA; i.e., the number of modern human template DNA molecules in the PCR mixture therefore exceeded by far putative human DNA molecules from the spiking extracts by several orders of magnitudes. Second, human DNA from other sources than the targeted modern human DNA can hardly explain why 11 mutations were found in more than 30% of the analyzed clones (table 3). Third, human DNA from other sources than the targeted modern human DNA can hardly explain that a combination of, e.g., 16 mutations (interim consensus sequence XXVI in table 2) occurs in 86 clones, i.e., 15.7% of all sequences, generated in seven different spiking experiments. Fourth, human DNA from other sources than the targeted modern human DNA can hardly explain why many interim consensus sequences occur that do not match any HVRI sequences deposited in sequence databases. Finally, spiking experiments using six ancient DNA extracts pretreated with DNaseI (67 sequences) also yielded mutated sequences with similar nucleotide alterations. Accordingly, DNA could be excluded as the mutagenic agent.
We speculate that coextracted chemicals other than DNA might induce the observed sequence alterations. Such coextracted mutagenic substances may accumulate in the samples during diagenesis. Although diagenetic processes are barely understood, we know that specimens accumulate postmortem multivalent metal ions. For example, concentrations of manganese, barium, strontium, iron, and uranium can be enriched up to 5,000 times compared with contemporary samples (Bischoff 1981, Velasco-Vazquez et al. 1997, Kohn et al. 1999). In particular, manganese is known for its mutagenic effect, and molecular biologists use this effect when using mutagenesis kits. In preliminary studies, we noticed substantially elevated manganese concentrations for the samples used here (data not shown). In contrast, manganese was not detected in 87 modern controls. If coextracted manganese is the reason for the observed mutations, one would, at least at first glance, expect to see a predictive relationship between manganese concentration and reliability of sequences. However, there is no such relationship, and, even worse, there is no reason to expect such a simple correlation. This is because the chemistry of manganese is very complex. The element occurs with a variety of valence stages that differ in their biological activity. However, manganese measurements refer to the total amount of manganese and cannot differentiate valence stages. In addition, the mutagenic effect of manganese also depends on the concentration of other ions. Svetlov and Cooper (1998) and Kunichika et al. (2002), for example, noticed that the mutagenic effect of manganese changed with changing concentration of magnesium. Other reasons may be added (see, e.g., Fletcher et al. 1994, Pelletier et al. 1996). Nevertheless, Pusch et al. (in preparation) showed in other experiments that a GA transition occurs frequently at cDNA position 1,138 of the human FGFR3 gene in PCR amplifications under the presence of 0.25 mM MnCl2. The same diagnostic mutation also occurs when contemporary human template DNA is spiked with an ancient DNA extract from a Pleistocene Ursus spelaeus bone. Future research will have to address this topic.
If coextracted chemicals such as multivalent metal ions were the reason for the observed sequence alterations, they might be removable by extraction methods, which are based on different extraction principles than the default silica-based protocol. In preliminary experiments, we also tested two other extraction methods, i.e., the "voltage-induced release method" (Bachmann et al. 2000) and the "mix-and-clean method" (Scholz and Pusch 1997). However, DNA extracted according to these alternative methods also yielded altered sequences in spiking experiments (data not shown). Future studies will have to address why hypothetical coextracted mutagenic substances cannot be removed by currently applied methods.
The data presented here have implications for ancient DNA research and may challenge the authenticity of a number of PCR-generated sequences from ancient samples. For some reasons it is, however, impossible to assess how frequently sequence alterations induced by ancient DNA extracts may have affected published sequences. Many papers report only the deduced consensus sequences and neither the sequences of the individual clones nor information on the observed sequence variation. Furthermore, as can be seen from table 1, different ancient DNA extracts induce sequence alterations to a different extent. For some extracts, the majority of sequences were identical to CRS and the obtained mutations occurred at very low frequencies. In those cases, extract-induced sequence alterations would not make it into a consensus sequence when applying the majority rule (hypothetical consensus sequence A in tables 1 and 4). In other instances, the majority of sequences were altered and the deduced consensus sequences wrong (hypothetical consensus sequences B–J in tables 1 and 4). It might also be a matter of chance which mutations are considered to be authentic, because usually the clones used for analysis are selected at random.
Table 4 Deduced Hypothetical Consensus Sequences of Cloned PCR Products of a 148-bp Stretch of Mitochondrial HVRI (16147–16294) of a Contemporary Human Amplified Under the Presence of Various Ancient DNA Extracts (table 1) (CRS is the correct sequence).
Intuitively, one might feel that applying the rule to clone and sequence short overlapping PCR products from several independent amplifications in order to deduce a consensus sequence may prevent from pitfalls. According to our data, this assumption is not justified. In the data set presented here 11 alterations occurred in more than 30% of the analyzed clones and some of those would make it into a consensus sequence in an ancient DNA approach when following the guidelines for ancient DNA. Independent replication of the experiments in other laboratories to verify deduced consensus sequences of ancient samples are of no help since such mutations can, as shown, repeatedly occur in independent amplifications. As stated earlier, particular positions have a high likelihood of being altered.
Our data imply that an unknown number of published ancient DNA sequences have to be considered with caution. There is an urgent need to identify the causes for the observed mutations in the spiking experiments and to develop standards for an assessment of the mutagenic effect of ancient DNA extracts as a mandatory criterion for authenticity. Although to a different extent in different studies, it is likely that extract-induced mutations are an intrinsic and a general problem of PCR amplifications of ancient templates.
Acknowledgements
We are grateful to M. Broghammer for excellent technical assistance. This work was supported in part by grants from the Research Council of Norway (National Centre for Biosystematics, 146515/420), the German Federal Ministry of Education, Science, Research and Technology (F?. 01KS9602), and the Interdisciplinary Center of Clinical Research (IZKF-Q3) Tübingen.
Literature Cited
Anderson, S., A. T. Bankier, and B. G. Barrel, et al. (14 coauthors). 1981. Sequence and organisation of the human mitochondrial genome. Nature 290:457-465.
Bachmann, L., M. Scholz, M. Broghammer, I. Giddings, and C. M. Pusch. 2000. Voltage-induced release of nucleic acids from palaeontological samples. Electrophoresis 21:1488-1492.
Bischoff, J. L. 1981. A geochemical study of the Eyasi I skull: implications for the process of fossilization. Pp. 128–132 in: H. Müller-Beck, ed. The palaeoanthropological finds of the Pliocene and Pleistocene. Archaeologica Venatoria, Tübingen-Bad Bramstedt.
Boom, R., C. J. Sol, M. M. Salimans, C. L. Jansen, P. M. Wertheim-van Dillen, and J. van der Noordaa. 1990. Rapid and simple method for purification of nucleic acids. J. Clin. Microbiol. 28:495-503.
Caramelli, D., C. Lalueza-Fox, and C. Vernesi, et al. (11 coauthors). 2003. Evidence for a genetic discontinuity between Neandertals and 24,000-year-old anatomically modern Europeans. Proc. Natl. Acad. Sci. USA. 27:6593-6597.
Cooper, A., and H. N. Poinar. 2000. Ancient DNA: do it right or not at all. Science 289:1139.
Fletcher, G. G., F. E. Rossetto, J. D. Turnbull, and E. Nieboer. 1994. Toxicity, uptake, and mutagenicity of particulate and soluble nickel compounds. Environ. Health Perspect. 102:(Suppl.) 3: 69-79.
Gilbert, M.T., A. J. Hansen, E. Willerslev, L. Rudbeck, I. Barnes, N. Lynnerup, and A. Cooper. 2003a. Characterization of genetic miscoding lesions caused by postmortem damage. Am. J. Hum. Genet. 72:48-61.
Gilbert, M. T., E. Willerslev, A. J. Hansen, I. Barnes, L. Rudbeck, N. Lynnerup, and A. Cooper. 2003b. Distribution patterns of postmortem damage in human mitochondrial DNA. Am. J. Hum. Genet. 72:32-47 Erratum in: Am. J. Hum. Genet. 72:779.
Handt O., M. Richards, and M. Trommsdorff, et al. (13 coauthors). 1994. Molecular genetic analyses of the Tyrolean Ice Man. Science 264:1775-1778.
Hansen, A. J., E. Willerslev, C. Wiuf, T. Mourier, and P. Arctander. 2001. Statistical evidence for miscoding lesions in ancient DNA templates. Mol. Biol. Evol. 18:262-265.
Hofreiter, M., V. J?nicke, D. Serre, A. von Haeseler, and S. P??bo. 2001a. DNA sequences from multiple amplifications reveal artifacts induced by cytosine deamination in ancient DNA. Nucleic Acids Res. 29:4793-4799.
Hofreiter, M., D. Serre, H. N. Poinar, M. Kuch, and S. P??bo. 2001b. Ancient DNA. Nature Rev. Genet. 2:353-359.
H?ss, M., and S. P??bo. 1993. DNA extraction from Pleistocene bones by a silica-based purification method. Nucleic Acids Res. 21:3913-3914.
Kohn, M. J., M. J. Schoeninger, and W. W. Barker. 1999. Altered states: effects of diagenesis on fossil tooth chemistry. Geochim. Cosmochim. Acta 63:2737-2747.
Krings, M., A. Stone, R. W. Schmitz, H. Krainitzki, M. Stoneking, and S. P??bo. 1997. Neandertal DNA sequences and the origin of modern humans. Cell 90:19-30.
Kunichika, K., H. Hashimoto, and T. Imoto. 2002. Robustness of hen lysozyme monitored by random mutations. Protein Eng. 15:805-809.
Lindahl, T. 1996. The Croonian Lecture, 1996: endogenous damage to DNA. Philos. Trans. R. Soc. Lond. B Biol. Sci. 351:1529-1538.
P??bo, S. 1989. Ancient DNA: extraction, characterization, molecular cloning, and enzymatic amplification. Proc. Natl. Acad. Sci. USA 86:1939-1943.
Pelletier, H., M. R. Sawaya, W. Wolfle, S. H. Wilson, and J. Kraut. 1996. A structural basis for metal ion mutagenicity and nucleotide selectivity in human DNA polymerase beta. Biochemistry 35:12762-12777.
Poinar, H., M. Kuch, G. McDonald, P. Martin, and S. P??bo. 2003. Nuclear gene sequences from a late pleistocene sloth coprolite. Curr. Biol. 13:1150-1152.
Pusch, C. M. 1997. A simple and fast procedure for high quality DNA isolation from gels using laundry detergent and inverted columns. Electrophoresis 18:1103-1104.
Pusch, C. M., L. Bachmann, M. Broghammer, and M. Scholz. 2000. Internal Alu-polymerase chain reaction: a sensitive contamination monitoring protocol for DNA extracted from prehistoric animal bones. Anal. Biochem. 284:408-411.
Scholz, M., and C. Pusch. 1997. An efficient isolation method for high-quality DNA from ancient bones. Trends Genet. 13:249.
Svetlov, V., and T. G. Cooper. 1998. Efficient PCR-based random mutagenesis of sub-genic (100bp) DNA fragments. Yeast 14:89-91.
Velasco-Vazquez, J., M. Arnay-de-la-Rosa, E. Gonzalez-Reimers, and O. Hernandez-Torres. 1997. Paleodietary analysis on the prehistoric population of El Hierro (Canary Islands). Biol. Trace Elem. Res. 60:235-241.
Willerslev, E., A. J. Hansen, B. Christensen, J. P. Steffensen, and P. Arctander. 1999. Diversity of Holocene life forms in fossil glacier ice. Proc. Natl. Acad. Sci. USA 96:8017-8021.(Carsten M. Pusch* and Lut)