Ancient DNA Damage

Jesse Dabney, Matthias Meyer, and Svante Pääbo

Department of Evolutionary Genetics, Max Planck Institute for Evolutionary Anthropology,
04103 Leipzig, Germany
Correspondence: [email protected]

Under favorable conditions DNA can survive for thousands of years in the remains of dead
organisms. The DNA extracted from such remains is invariably degraded to a small average
size by processes that at least partly involve depurination. It also contains large amounts of
deaminated cytosine residues that are accumulated toward the ends of the molecules, as well
as several other lesions that are less well characterized.

n living cells, DNA molecules continuously fest themselves in three different ways: (i) a re-
I suffer chemical insults, which are countered
by enzymatic repair mechanisms that maintain
duction in DNA fragment size, (ii) lesions that
block the replication of the DNA molecules by
the integrity of the genome (Lindahl 1993). On polymerases, thus impeding many forms of
death, these cellular repair mechanisms cease to analysis, and (iii) lesions that cause incorrect
function. As a consequence, the genome be- nucleotides to be incorporated when the DNA
comes exposed to the unmitigated effects of nu- is replicated. Here, we summarize what is known
merous factors that threaten its stability. These about each of these forms of damage in ancient
factors include intracellular nucleases, which are DNA.
no longer sequestered in the cell and can thus
gain access to DNA and degrade it, as well as
FRAGMENTATION
microorganisms that spread in the decaying tis-
sues. Together these factors may lead to the loss Already in the first systematic study of the prop-
of all retrievable DNA. However, under favorable erties of ancient DNA (Pääbo 1989), it was
environmental conditions, for example when shown that almost all DNA extracted from sam-
tissues are frozen or become desiccated quickly ples varying in age between 4 and 13,000 years
after death, these processes become inhibited was degraded to fragments of 40 – 500 bp. Sub-
before the complete destruction of all DNA en- sequent work has confirmed that this is a general
dogenous to the organism. In these instances feature of DNA extracted from almost all ancient
other destructive factors, particularly hydrolytic remains. Based on in vitro experiments using
and oxidative processes, become limiting to the modern DNA (Lindahl and Andersson 1972;
time that DNA survives in a tissue. Lindahl and Nyberg 1972), it has been suggested
When DNA is extracted and analyzed from that fragmentation is owing to hydrolytic depu-
ancient samples these destructive factors mani- rination and subsequent b elimination resulting

Editors: Errol C. Friedberg, Stephen J. Elledge, Alan R. Lehmann, Tomas Lindahl, and Marco Muzi-Falconi
Additional Perspectives on DNA Repair, Mutagenesis, and Other Responses to DNA Damage available at www.cshperspectives.org
Copyright # 2013 Cold Spring Harbor Laboratory Press; all rights reserved; doi: 10.1101/cshperspect.a012567
Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567

1
J. Dabney et al.

in single-strand breaks (Fig. 1) (Pääbo and Wil- and not removed by enzymatic treatments
son 1991; Lindahl 1993). Indeed, alkali treat- used during library preparation. Using this ap-
ment of ancient DNA extracts indicates the pres- proach and a reference genome to infer the bas-
ence of terminal 50 -phosphate and aldehydic es immediately adjacent to the ancient DNA
30 ends, which are products of b elimination fragments, it was shown that the purines ade-
(Jones et al. 1968). nine (A) and guanine (G) are overrepresented
During the last decade, high-throughput se- next to the 50 ends of DNA fragments extracted
quencing methods have allowed a better under- from the remains of Neanderthals, mammoths,
standing of fragmentation patterns in ancient and cave bears, which were all 40,000 years old
DNA. This is because these methods rely on (Fig. 2) (Briggs et al. 2007). In Pleistocene horse
the ligation of DNA adaptors to the ends of remains preserved in permafrost (Orlando et al.
DNA molecules to construct sequencing librar- 2011) it was shown that G residues were even
ies (Margulies et al. 2005). Subsequent sequenc- more overrepresented than A residues adjacent
ing from the adaptors allows the location of to the 50 ends. A possible explanation for the
strand breaks to be determined as long as the preferential depurination of G over A residues
terminal nucleotides are amenable to ligation is a resonance structure in guanine that may

A O
Purine
N (guanine)
O– O–
N
O P O O P O
O– OH OH
5′ N N NH2 C O Abasic C O
O– O–
O P O site
O Hydrolysis
O–
DNA sugar O O–
H+, H2O OH–
phosphate
backbone 3′ Depurination O P O β Elimination O P O
O
O– O–
O P O

O–

B Amine
Carbonyl group
group

NH2 O
Hydrolysis

+ H2O
N NH

N N
O O
H H

Cytosine Uracil

Figure 1. Fragmentation and deamination. (A) A likely cause of fragmentation in ancient DNA is depurination,
in which the N-glycosyl bond between a sugar and an adenine or guanine residue is cleaved, resulting in an abasic
site. The DNA strand is then fragmented through b elimination, leaving 30 -aldehydic and 50 -phosphate ends. (B)
Deamination of cytosine to uracil is the major mechanism leading to miscoding lesions in ancient DNA. DNA
polymerases will incorporate an A across from the U, and in turn a Tacross from the A, causing apparent G to A
and C to T substitutions.

2 Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567

 Ancient DNA Damage

5′ Reference sequence fragments in DNA extracted from bones pre-

0.35 served for tens of thousands of years in caves
5′ Neanderthal sequence
(Meyer et al. 2012).
0.30
Base frequency BLOCKING LESIONS
0.25
Some DNA modifications obstruct the move-
ment of DNA polymerases along a template
0.20
strand, preventing their amplification and se-
A quencing. Such blocking lesions occur in the
G
0.15
C form of nucleotide modifications and cross-
T
links, which can form either between DNA
0.10 strands, between different DNA fragments, or
−10 −5 0 5 10 between DNA and other molecules.
Nucleotide position from 5′ end Several nucleotide modifications block po-
Figure 2. Base frequencies of the human reference
lymerase-mediated DNA synthesis. In an anal-
genome around 50 ends of Neanderthal DNA frag- ysis of DNA extracted from 11 samples repre-
ments. An excess of A and G immediately outside of senting various permafrost and nonpermafrost
the Neanderthal sequence indicates that an elevated preservation conditions, it was shown that dif-
number of DNA fragments begin adjacent to purines, ferent amounts of 5-hydroxy-5-methylhydan-
supporting the hypothesis that depurination is a ma- toin and 5-hydroxyhydantoin, which are oxi-
jor mechanism of fragmentation of ancient DNA. dation products of pyrimidines, were present
in all samples. Amplification of mitochondri-
lower the activation energy required to break al DNA was only possible in five of the sam-
its N-glycosyl bond (Overballe-Petersen et al. ples that contained threefold to 29-fold less hy-
2012). dantoins than samples in which amplification
Most methods for preparing DNA libraries failed (Höss et al. 1996). Little is known about
for sequencing involve blunt-end repair, an en- other blocking lesions in ancient DNA although
zymatic process that extends recessed and de- primer extension experiments have shown an
grades overhanging 30 ends of DNA fragments, elevated termination frequency at guanine resi-
thus precluding the determination of the ex- dues in ancient samples (Heyn et al. 2010), sug-
act location of 30 ends (Briggs et al. 2007). A gesting that a modification of guanine may also
first attempt to determine the sequence con- act as a blocking lesion.
text around 30 ends was made using a sequenc- Early work using electron microscopy iden-
ing technology that sequences from polyad- tified intermolecular cross-links in ancient DNA
enylated, but otherwise unmodified, 30 ends (Pääbo 1989). In addition, it has been shown by
(Orlando et al. 2011). For permafrost-preserved gas chromatography/mass spectrometry (GC/
Pleistocene horse DNA this yielded no indica- MS) that furanones, furaldehydes, and alkyl-
tion that strand breakage occurs preferentially pyrazines—products of the Maillard reaction,
at any particular nucleotide. However, this may which can produce cross-links between macro-
be owing to inefficient poly(A)-tailing of alde- molecules like DNA and proteins—are present
hydic 30 ends, which would exclude fragments in 20,000-yr-old sloth coprolites (Poinar et al.
generated by b elimination (Orlando et al. 1998). Amplification of a short mitochondri-
2011). A recently developed DNA library prep- al DNA fragment from the coprolites was pos-
aration method in which adaptors are ligated to sible only after treatment with N-phenacylthia-
single-stranded DNA, preserves both the 50 and zolium bromide (N-PTB), a compound that
30 ends and has confirmed that purines, espe- cleaves Maillard products. However, although
cially guanine residues, are overrepresented im- N-PTB is helpful in some instances, subsequent
mediately adjacent to both ends of ancient DNA studies have reported less success using N-PTB

Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567 3

J. Dabney et al.

treatments to improve amplification yield (Roh- number over other substitutions (Stiller et al.
land and Hofreiter 2007; Binladen and Willer- 2006; Gilbert et al. 2007). Moreover, it has been
slev 2010). found that these substitutions are primarily lo-
It has also been suggested, based on the re- calized to fragment ends where up to 40% of all
fractivity of permafrost-derived ancient DNA cytosines appear as thymines, followed by an
to heat denaturation, that cross-links accumu- exponential decrease of such substitutions along
late about 100 times faster than single-strand the DNA molecule (Fig. 3) (Briggs et al. 2007;
breaks in such DNA (Hansen et al. 2006). In Brotherton et al. 2007). Notably, blunt-end re-
contrast, primer extensions on ancient DNA pair by T4 DNA polymerase used to prepare the
fragments that were 27,000 – 48,000 years old DNA sequencing libraries results in the elimina-
and came from three bones preserved in perma- tion of overhanging 30 ends, whereas overhang-
frost and one bone from a cave site concluded ing 50 ends are filled in. Consequently, toward the
that all blocking lesions, including cross-links, 30 ends of the sequenced molecules, deaminated
were present in no more than 40% of molecules cytosine residues result in apparent G to A sub-
(Heyn et al. 2010). Thus, there is currently con- stitutions, explaining the occurrence of these
flicting evidence about the extent to which substitutions in libraries prepared from ancient
blocking lesions are present in ancient DNA. DNA (Briggs et al. 2007; Brotherton et al. 2007),
which was initially mistaken for deamination of
guanine residues (Stiller et al. 2006). Analysis of
MISCODING LESIONS
the 30 end of ancient DNA fragments, made pos-
Nucleotide bases are susceptible to hydrolytic sible by both new sequencing technologies (Or-
deamination, resulting in modifications that lando et al. 2011) and the recent single-stranded
cause them to be misread by DNA polymerases. library preparation method (Meyer et al. 2012),
A primary target of deamination is cytosine. Its unequivocally shows that C to T changes occur at
product, uracil (Fig. 1), will direct the incorpo-
ration of adenine (A) during DNA replication,
resulting in apparent C to Tor G to A substitu-
tions (depending on the strand sequenced). Be-
C to T
cause ancient DNA is sensitive to uracil-N-gly- 0.4 Other
cosylase (UNG) it has been inferred that it
contains uracil residues (Pääbo 1989). Further-
Substitution frequency

0.3
more, asymmetric PCR that amplifies only one
DNA strand has shown that the large majority of
nucleotide substitutions in ancient DNA se- 0.2
quences are from C to T and that treatment
with UNG results in a substantial reduction of
such substitutions (Hofreiter et al. 2001). Some 0.1

investigators have proposed that adenine is de-

aminated to hypoxanthine in ancient DNA, 0.0
causing apparent A to G changes during ampli- 0 5 10 15 20
fication (Hansen et al. 2001; Gilbert et al. 2003). Nucleotide position from 5′ end
Although this may occur at some level, many of
these substitutions are likely to be the result of Figure 3. Nucleotide substitution frequencies at the 50
damage-independent misincorporations by the end of Neanderthal sequences. The frequencies of all
possible differences to the human reference genome
Taq polymerase used for PCR (Hofreiter et al.
sequence are plotted in relation to their location in
2001; Pääbo et al. 2004; Binladen et al. 2006). Neanderthal DNA sequences. Cytosine to thymine
High-throughput sequencing of libraries substitutions are the most frequent, and they occur
constructed from ancient DNA has confirmed most often at ends of molecule. All remaining nucle-
that C to T substitutions are greatly increased in otide substitutions are indicated by the black line.

4 Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567

 Ancient DNA Damage

similar and high frequencies on both ends of (Willerslev et al. 2007). In contrast, a number
ancient DNA fragments. of publications of DNA sequences determined
Because deaminated cytosines are localized from remains that are several million years old
primarily to the ends of DNA fragments and exist in the literature, including inclusions of
because the rate of cytosine deamination is animals and plants enclosed in amber (DeSalle
about two orders of magnitude faster in sin- et al. 1992) and fossils of plants and dinosaurs
gle-stranded than double-stranded DNA (Lin- (Golenberg et al. 1990; Woodward et al. 1994).
dahl 1993), it is likely that the accumulation These probably all represent DNA from present-
of C to T substitutions at the ends of ancient day organisms that have contaminated the sam-
DNA fragments reflects the occurrence of sin- ples or experiments (Pääbo and Wilson 1991;
gle-stranded overhanging ends in ancient DNA Lindahl 1993; Pääbo et al. 2004).
(Briggs et al. 2007; Orlando et al. 2011). Because DNA sequence features indicative of DNA
uracils occur only rarely in the middle of ancient damage have recently been assessed in animal
DNA strands they can be removed using uracil- remains varying in age from 18 to 60,000 years
N-glycosylase and Escherichia coli endonuclease (Sawyer et al. 2012). No correlation between
VIII to improve sequence quality without sub- fragment length and age was found, and only a
stantially impairing the overall recovery of DNA weak negative correlation between age and the
sequence (Briggs et al. 2010). occurrence of purines immediately adjacent
to the 50 ends of fragments was seen. Interest-
ingly, in samples younger than 100 years of age,
DAMAGE THROUGH TIME
adenine residues predominated adjacent to the
Although environmental conditions such as hu- 50 ends of fragments, whereas guanine residues
midity, temperature, salinity, and pH will have a predominated in samples older than 40,000
strong effect on DNA preservation, it has been years. This suggests that some process other
estimated through extrapolation from in vitro than depurination, which likely causes the ele-
experiments (Lindahl 1993) that DNA would vated guanine signal in older samples, occurs in
survive no more than a few hundred thousand younger samples, perhaps degradation by en-
years. Indeed, it was recently estimated that the zymes shortly after death. Of the DNA sequence
half-life of 242 bp mitochondrial DNA frag- features studied, only cytosine deamination was
ments in bird bones excavated in a small area strongly positively correlated with age, despite
in New Zealand is around 500 years and that the fact that the samples came from different
only preservation in frozen conditions might sites and a variety of burial conditions. Cytosine
allow DNA to survive for more than a million deamination, manifested by an elevated C to T
years (Allentoft et al. 2012). When several such substitution frequency, can therefore be used as
detailed studies of DNA survival from different an indication that DNA molecules are indeed
environments become available it will hopefully ancient (Krause et al. 2010).
be possible to estimate the likelihood of DNA
survival given the environmental conditions at
CONCLUDING REMARKS
an archaeological site. However, even then, con-
ditions of relevance for DNA survival such as Our understanding of postmortem DNA dam-
amount of water percolation, salinity, pH, and age is still very fragmentary. For example, de-
microbial growth are likely to vary within a site purination is a well-characterized mechanism
or an archaeological stratum and make preser- of DNA degradation but its contribution to an-
vation conditions variable on a “microscale.” cient DNA fragmentation is based entirely on
Nevertheless, it is clear that DNA will not indirect evidence and it probably accounts for
survive over geological time scales. The oldest only 10%– 40% of the fragmentation that oc-
credible samples that have been sequenced stem curs in ancient DNA (Sawyer et al. 2012). Fur-
from plant and insect DNA found in 450,000- thermore, it is unclear how prevalent block-
to 800,000-yr-old ice cores from Greenland ing lesions are in ancient DNA. Fortunately,

Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567 5

J. Dabney et al.

technologies are likely to emerge that will allow 2007. Recharacterization of ancient DNA miscoding le-
sions: Insights in the era of sequencing-by-synthesis. Nu-
nucleotide modifications and lesions to be di- cleic Acids Res 35: 1 –10.
rectly detected and characterized without prior Golenberg EM, Giannasi DE, Clegg MT, Smiley CJ, Dur-
amplification or enzymatic modifications (e.g., bin M, Henderson D, Zurawski G. 1990. Chloroplast
Branton et al. 2008). An improved understand- DNA sequence from a miocene Magnolia species. Nature
344: 656–658.
ing of ancient DNA damage may allow the devel-
Hansen AJ, Willerslev E, Wiuf C, Mourier T, Arctander P.
opment of new repair strategies that might in- 2001. Statistical evidence for miscoding lesions in ancient
crease the number of DNA sequences that can DNA templates. Mol Biol Evol 18: 262 –265.
be retrieved from ancient remains even further. Hansen AJ, Mitchell DL, Wiuf C, Paniker L, Brand TB,
Binladen J, Gilichinsky D A, Rønn R, Willerslev E.
2006. Crosslinks rather than strand breaks determine ac-
ACKNOWLEDGMENTS cess to ancient DNA sequences from frozen sediments.
Genetics 173: 1175– 1179.
We thank Susanna Sawyer and the Multimedia Heyn P, Stenzel U, Briggs AW, Kircher M, Hofreiter M,
Meyer M. 2010. Road blocks on paleogenomes–poly-
Department for assistance with figures. Our merase extension profiling reveals the frequency of block-
work is funded by the Max Planck Society. ing lesions in ancient DNA. Nucleic Acids Res 38: e161.
Hofreiter M, Jaenicke V, Serre D, von Haeseler A, Pääbo S.
2001. DNA sequences from multiple amplifications re-
REFERENCES veal artifacts induced by cytosine deamination in ancient
DNA. Nucleic Acids Res 29: 4793–4799.
Allentoft ME, Collins M, Harker D, Haile J, Oskam CL, Höss M, Jaruga P, Zastawny TH, Dizdaroglu M, Pääbo S.
Hale ML, Campos PF, Samaniego JA, Gilbert MTP, 1996. DNA damage and DNA sequence retrieval from
Willerslev E, et al. 2012. The half-life of DNA in bone: ancient tissues. Nucleic Acids Res 24: 1304– 1307.
Measuring decay kinetics in 158 dated fossils. Proc R
Soc B doi: 10.1098/rspb.2012.1745. Jones AS, Mian AM, Walker RTJ. 1968. Chem Soc C 2042–
2044.
Binladen J, Willerslev E. 2010. Why study ancient DNA
damage? J Nordic Archaeol Sci 14: 11–14. Krause J, Briggs AW, Kircher M, Maricic T, Zwyns N,
Derevianko A, Pääbo S. 2010. A complete mtDNA ge-
Binladen J, Wiuf C, Gilbert MTP, Bunce M, Barnett R, Lar- nome of an early modern human from Kostenki, Russia.
son G, Greenwood AD, Haile J, Ho SYW, Hansen AJ, et al. Curr Biol 20: 231 –236.
2006. Assessing the fidelity of ancient DNA sequences
amplified from nuclear genes. Genetics 172: 733– 741. Lindahl T. 1993. Instability and decay of the primary struc-
ture of DNA. Nature 362: 709–715.
Branton D, Deamer DW, Marziali A, Bayley H, Benner SA,
Butler T, Di Ventra M, Garaj S, Hibbs A, Huang X, et al. Lindahl T, Andersson A. 1972. Rate of chain breakage at
2008. The potential and challenges of nanopore se- apurinic sites in double-stranded deoxyribonucleic
quencing. Nat Biotechnol 26: 1146– 1153. acid. Biochemistry 11: 3618– 3623.
Briggs AW, Stenzel U, Johnson PLF, Green RE, Kelso J, Lindahl T, Nyberg B. 1972. Rate of depurination of native
Prüfer K, Meyer M, Krause J, Ronan MT, Lachmann M, deoxyribonucleic acid. Biochemistry 11: 3610– 3618.
et al. 2007. Patterns of damage in genomic DNA sequenc- Margulies M, Egholm M, Altman WE, Attiya S, Bader JS,
es from a Neandertal. Proc Natl Acad Sci 104: 14616– Bemben LA, Berka J, Braverman MS, Chen Y-J, Chen Z,
14621. et al. 2005. Genome sequencing in microfabricated high-
Briggs AW, Stenzel U, Meyer M, Krause J, Kircher M, density picolitre reactors. Nature 437: 376 –380.
Pääbo S. 2010. Removal of deaminated cytosines and Meyer M, Kircher M, Gansauge M-T, Li H, Racimo F,
detection of in vivo methylation in ancient DNA. Nucleic Mallick S, Schraiber JG, Jay F, Prüfer K, De Filippo C,
Acids Res 38: e87. et al. 2012. A high-coverage genome sequence from an
Brotherton P, Endicott P, Sanchez JJ, Beaumont M, Bar- archaic Denisovan individual. Science (New York, NY)
nett R, Austin J, Cooper A. 2007. Novel high-resolution 338: 222–226.
characterization of ancient DNA reveals C . U-type Orlando L, Ginolhac A, Raghavan M, Vilstrup J, Rasmu-
base modification events as the sole cause of post mortem ssen M, Magnussen K, Steinmann KE, Kapranov P,
miscoding lesions. Nucleic Acids Res 35: 5717– 5728. Thompson JF, Zazula G, et al. 2011. True single-molecule
DeSalle R, Gatesy J, Wheeler W, Grimaldi D. 1992. DNA DNA sequencing of a pleistocene horse bone. Genome Res
sequences from a fossil termite in Oligo-Miocene amber 21: 1705–1719.
and their phylogenetic implications. Science (New York, Overballe-Petersen S, Orlando L, Willerslev E. 2012. Next-
NY) 257: 1933–1936. generation sequencing offers new insights into DNA deg-
Gilbert MTP, Willerslev E, Hansen AJ, Barnes I, Rudbeck L, radation. Trends Biotechnol 30: 364 –368.
Lynnerup N, Cooper A. 2003. Distribution patterns of Pääbo S. 1989. Ancient DNA: Extraction, characterization,
postmortem damage in human mitochondrial DNA. Am molecular cloning, and enzymatic amplification. Proc
J Hum Genet 72: 32–47. Natl Acad Sci 86: 1939– 1943.
Gilbert MTP, Binladen J, Miller W, Wiuf C, Willerslev E, Pääbo S, Wilson AC. 1991. Miocene DNA sequences—A
Poinar H, Carlson JE, Leebens-Mack JH, Schuster SC. dream come true? Curr Biol 1: 45– 46.

6 Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567

 Ancient DNA Damage

Pääbo S, Poinar H, Serre D, Jaenicke-Despres V, Hebler J, Stiller M, Green RE, Ronan M, Simons JF, Du L, He W,
Rohland N, Kuch M, Krause J, Vigilant L, Hofreiter M. Egholm M, Rothberg JM, Keates SG, Keats SG, et al.
2004. Genetic analyses from ancient DNA. Annu Rev Ge- 2006. Patterns of nucleotide misincorporations during
net 38: 645– 679. enzymatic amplification and direct large-scale sequenc-
Poinar HN, Hofreiter M, Spaulding GW, Martin PS, Stan- ing of ancient DNA. Proc Natl Acad Sci 103: 13578–
kiewicz BA, Bland H, Evershed RP, Possnert G, Pääbo S. 13584.
1998. Molecular coproscopy: Dung and diet of the extinct
Willerslev E, Cappellini E, Boomsma W, Nielsen R,
ground sloth Nothrotheriops shastensis. Science 281:
Hebsgaard MB, Brand TB, Hofreiter M, Bunce M,
402–406.
Poinar HN, Dahl-Jensen D, et al. 2007. Ancient bio-
Rohland N, Hofreiter M. 2007. Comparison and optimiza-
tion of ancient DNA extraction. BioTechniques 42: 343 – molecules from deep ice cores reveal a forested
352. southern Greenland. Science (New York, NY) 317: 111–
Sawyer S, Krause J, Guschanski K, Savolainen V, Pääbo S. 114.
2012. Temporal patterns of nucleotide misincorporations Woodward SR, Weyand NJ, Bunnell M. 1994. DNA sequence
and DNA fragmentation in ancient DNA. PloS ONE 7: from Cretaceous period bone fragments. Science (New
e34131. York, NY) 266: 1229–1232.

Cite this article as Cold Spring Harb Perspect Biol 2013;5:a012567 7