The fragmentised remains (tibia, skulls, jaw and maxilla) of a Cro-Magnon individual, named Paglicci 23, were excavated by F.M in 2003 from the Paglicci cave, Southern Italy. Radiocarbon tests dated the layer to 28.100 (+/−350) years ago 
. Because of its fragmentary nature, the sample was neither restored nor studied from the morpho-anatomical point of view. Therefore, no contamination could possibly be introduced at these stages by direct handling and washing. The remains were deposited in the storage room at controlled temperature in the Department of Archaeology, University of Pisa. In 2005 three splinters, a piece of tibia () and two pieces of skull, were moved to the ancient DNA laboratory at the University of Florence. In the course of the whole process, from excavation of the remains to genetic typing, only seven persons had any contacts with the sample, namely six of us (F.M, S.V, A.M., E.Pi., M.L., and D.C.) and Carles Lalueza-Fox (hereafter: C.L.) who replicated the sequence at the University Pompeu Fabra, Barcelona.
Tibia fragment of the Paglicci 23 specimen.
The degree of racemization of three amino acids, aspartic acid, alanine, and leucine, provides indirect evidence as for the presence in an ancient sample of amplifiable DNA. In particular, DNA is expected to be too degraded for amplification when the D/L for Asp is greater than 0.08 
. As a preliminary test of macromolecule preservation, we measured the stereoisomeric D/L ratio for these amino acids. The observed values, all of them compatible with good preservation of biological macromolecules in the sample, were D/L Asp 0.0479, D/L Glu 0.0104 D/L, Ala 0.0092. The global amino acid content was 42,589 parts per million, and endogenous DNA was successfully ampified from Pleistocene remains when this value was higher than 30,000 parts per million 
Quantitative PCR showed a relatively large amount of mtDNA molecules in the Paglicci 23 fossil, approximately 2300. Contamination, usually detected when different sequences are observed in different cloned products, is considered unlikely if the number of PCR template molecules is >1,000 
. We thus proceeded in the analysis by initially sequencing a total of 144 clones (Table S1
), respectively 64, 32 and 48 for the three regions in which the HVR I was divided. Reproducible mtDNA sequences corresponding to positions 16024–16383 of the published reference sequence CRS 
were obtained in the Florence laboratory from the tibia and from a skull fragment of Paglicci 23. No contamination was observed in the extractions and PCR blanks. Amplification of long DNA fragments, unusual for ancient DNA, was not observed. The analysis was repeated in Barcelona, using a tibia fragment; the consensus sequence obtained from 8 clones covering the region between nt 16245 to nt 16349 was identical to that obtained in Florence. On the contrary, no PCR product was observed when we attempted to amplify the DNA extracts using two pairs of Neandertal-specific primers.
As is common in studies of ancient DNA, when comparing sequences across clones we observed single nucleotide substitutions occurring in one or a few clones (Table S1
), on average 3.9 every 1,000 bp. In addition, a C to T change was observed in 27 out of 56 clones at nt 16274. In principle, differences of this kind across clones may be due to three factors, namely: (1) sequence heterogeneity due to the presence of exogenous, contaminating DNA, (2) post-mortem DNA damage, and (3) Taq
-polymerase errors or cloning artefacts. We tested separately for the possible effects of the first two factors upon our specimen.
To track down any possible modern contaminations, the mtDNAs of the seven authors who to any extent manipulated the sample were genotyped. All these sequences () differ from the Paglicci 23′s consensus mtDNA sequence. However, two of (F.M and C.L) have a T at nt 16274. Therefore, variation across clones at that site might have meant that either investigator left his DNA on the sample, although C.L. had no contacts with the material at the stage at which clones F2.1 through F2.13 and F3.1 through F3.15 were genotyped.
Mitochondrial HVR1 variation in the seven researchers that have been in physical contact with the samples.
Post-mortem DNA damage generally occurs in the form of double-strand breaks, or other modifications severe enough to prevent enzymatic replication of the DNA molecule. Had this happen, we would have been unable to amplify the DNA. However, hydrolytic deamination and depurination may also occur, resulting in apparent changes of the nucleotide sequence. Although post-mortem damages of this kind are unlikely to severely bias the results when the initial template molecules exceed 1000 
as is the case for Paglicci 23, to correct for such possible post-mortem damages, a third DNA extract was treated with Uracyl-N-Glycosidase (UNG) 
, and independently resequenced. The 35 sequences thus obtained (clones F 4.1 through F4.20, and F5.1 through F 5.15) contain no nucleotide substitutions with respect to the CRS, including nt 16274 (Table S1
). As a consequence, we concluded that the sequence obtained from the Paglicci 23 specimen is the CRS, and that heterogeneity across clones at nt 16274 reflects DNA damage due to deamination of the original cytosine and successive amplification of the damaged DNA fragment(s). The rate of nucleotide misincorporation suggests that the DNA templates were indeed damaged (3.9 substitutions every 1,000 bp within the HVRI), but after UNG treatment at least 82% of the clones showed the same consensus nucleotide at each position (Table S1
The relationship between the Paglicci 23 sequence, the available Cro-Magnon and Neandertal sequences, and all the sequences from the seven individuals who manipulated the Cro-Magnons specimen, are summarized in . The backbone of the network is based on the 31bp region for which we had complete overlap among all sequences, and was estimated by a statistical parsimony method 
, as implemented in the software TCS 
. A sub-network was also reconstructed for a set of eight individuals relevant to this study using the entire fragment of 360 bp.
Genetic relationships among the Paglicci 23 and other relevant mtDNA sequences.
Previous genetic data on Cro-Magnoids 
, although generated under the most stringent available criteria, were considered problematic by some authors 
, because the mtDNA sequences obtained correspond to sequences also observed in modern individuals. For most ancient human samples, rigorous application of this criterion would render the study of Cro-Magnoid DNA practically impossible, because it is impossible to rule out any contamination from generic unknown individuals. However, it is possible to test for the occurrence in the extract of known potential contaminating sequences; for the Paglicci 23 fossil we had this opportunity, and we found that none of these modern sequences is equal to the sequence obtained from the fossil extracts. Since we used different sets of overlapping primers pairs to amplify the fragment included between nucleotide 16024 and 16383, it seems highly unlikely that the sequence obtained was a chimera artefact.
Therefore, at this stage it is safe to conclude that at least one Cro-Magnoid mtDNA sequence, for which contamination can be ruled out with a high degree of confidence, falls well within the range of modern human variation. This does not prove, but at least indirectly suggests, that the previously published Cro-Magnoid sequences 
, both documented in the modern human gene pool, may be genuine 
. At any rate, the finding of the Cambridge Reference Sequence in Paglicci 23 shows that one of today's mtDNA variants has been present in Europe for at least 28,000 years, and that modern and archaic anatomical features appear associated with mtDNA sequences that can be classified, respectively, as modern and non-modern. Because no HVR I sequence similar to the Neandertals' has been described in more than 4800 Europeans studied so far 
, models whereby Neandertals were part of the genealogy of current Europeans are at odds with the data, at least as far as maternal inheritance is concerned. In our opinion, the burden of the proof is now on those who maintain that Neandertals might have contributed to the modern gene pool.
So far, the study of ancient nuclear DNA in humans has been severely limited by the difficulty to ascertain whether the DNA sequences obtained are really endogenous to the specimen. This study shows that it is possible to test for DNA authenticity, provided the people who manipulate the sample from the moment of excavation are carefully recorded and their DNAs typed. Therefore, Paglicci 23 (as well as other remains studied under comparable conditions in the future) promises to be a valuable source of information on DNA diversity in the past, and can pave the ground for a more exhaustive understanding of human evolutionary history.