SHORT COMMUNICATION

A Blind Testing Design for Authenticating Ancient DNA SequencesHong Yang,*,1 Edward M. Golenberg,* and Jeheskel Shoshani*,†

* Department of Biological Sciences, Wayne State University, Detroit, Michigan 48202; and †Cranbrook Institute of Science,Bloomfield Hills, Michigan 48304

Received August 13, 1996; revised December 10, 1996

Reproducibility is a serious concern among research-ers of ancient DNA. We designed a blind testing proce-dure to evaluate laboratory accuracy and authenticityof ancient DNA obtained from closely related extantand extinct species. Soft tissue and bones of fossil andcontemporary museum proboscideans were collectedand identified based onmorphology by one researcher,and other researchers carried out DNA testing on thesamples, which were assigned anonymous numbers.DNAextracted using three principal isolationmethodsserved as template in PCR amplifications of a segmentof the cytochrome b gene (mitochondrial genome), andthePCRproductwas directly sequenced and analyzed.The results show that such a blind testing designperformed in one laboratory,when coupledwith phylo-genetic analysis, can nonarbitrarily test the consis-tency and reliability of ancient DNA results. Suchreproducible results obtained from the blind testingcan increase confidence in the authenticity of ancientsequences obtained from postmortem specimens andavoid bias in phylogenetic analysis. A blind testingdesign may be applicable as an alternative to confirmancient DNA results in one laboratory when indepen-dent testing by two laboratories is not available. r 1997

Academic Press

Recent studies of ancient DNAfrom fossil and archeo-logical material have revealed unique insights into theevolutionary history of animals and plants, and thegenetic information obtained from ancient material isbeyond what can be extrapolated from living organ-isms. However, the criteria for establishing authentic-ity of ancient sequences remain a major concern amongresearchers of ancient DNA, and a focal point inestablishing authenticity of ancient DNA sequence has

been whether ancient DNA results can be reproduced(Lindahl, 1993; Handt et al., 1994a; Golenberg, 1994a;Richards et al., 1995). Efforts have been made toachieve reproducibility by amplifying different genesfrom the same sample (Hanni et al., 1994a; Beraud-Colomb et al., 1995) and by sequencing the same genesequence in different laboratories (Handt et al., 1994b;Hoss et al., 1996; Taylor, 1996). However, independenttests by two separate laboratories may not always bepossible, and even if such testings are performed, theyonly eliminate the possibility of laboratory-specificcontaminations. If the sample were contaminated be-fore reaching the two laboratories (e.g., during fieldcollection), identical but nonauthentic sequences wouldbe obtained. In addition, due to the difficulties involvedin ancient DNA experiments and different techniquesused between laboratories, the results may or may notalways be reproduced (Hanni et al., 1994b).Here, we report a blind experimental design that was

used in our recent investigations on molecular phylog-eny of ancient proboscideans (elephants and theirrelatives) as an alternative to confirmation of ancientDNA results in two laboratories. The blind testing canbe adequately performed in one laboratory, and it iseffective in monitoring laboratory techniques and inobjectively reproducing ancient DNA results. Such ablind testing design increases confidence in acceptingancient DNA sequences and reduces bias during se-quence analysis. We believe that if a blind testing isappropriately designed, it can produce reproducibleevidence to argue for authentication of ancient DNAsequence. We describe the design by usingmuseum andfossil animal specimens, but the results should give ageneral implication for other types of ancient materialsand forensic samples.

MATERIALS AND METHODS

Contemporary museum samples and fossil materialsof proboscideans, ranging in age from 2 years to over46,000 years old, were used in this study. Skin samples

1 To whom correspondence should be addressed at present address:Department of Human Genetics, University of Michigan MedicalSchool, Medical Science II M4708, Ann Arbor, MI 48109. Fax: (313)763 3784. E-mail: hongyang@umich.edu.

MOLECULAR PHYLOGENETICS AND EVOLUTION

Vol. 7, No. 2, April, pp. 261–265, 1997ARTICLE NO. FY960398

2611055-7903/97 $25.00Copyright r 1997 by Academic PressAll rights of reproduction in any form reserved.

were obtained from three extant elephants: Elephasmaximus (Asian elephant, EL 1, 14 years old) (Sho-shani et al., 1982), Loxodonta africana (African el-ephant, EL 4, 2 years old) (Shoshani, unpublisheddata), and a putative intergeneric hybrid elephant (EL3, 16 years old) between a male African elephant and afemale Asian elephant (Lowenstein and Shoshani,1996). A muscle specimen (EL 5) collected from thesame Asian elephant specimen (EL 1) was included forcomparison. Air-dried skin from a frozen carcass of anextinct Mammuthus primigenius (woolly mammothfrom Siberia, EL 2, more than 46,000 years old) (Sho-shani, unpublished data) was also used. Samples EL 1and EL 4 were preserved in salt, whereas samples EL 2and EL 5 were air dried. Bones from two extinctspecies, M. primigenius (woolly mammoth) and Mam-mut americanum (American mastodon), were used.Among them, one woolly mammoth sample (EL 19,20,000 years old) was collected from Alaska and thenwas stored in a museum (J. P. Alexander personalcommunication, 1994). Two rib fragments (EL 23 andEL 32) came from a 10,200-year-old American mast-odon skeleton which had yielded ancient proteins in aprevious study (Shoshani et al., 1985). Two additionalmastodon samples (EL 6, Shoshani et al., 1989 and EL29, Shoshani, unpublished data) were collected fromcranial remains unearthed from two Pleistocene fossilsites in Michigan.The blind testing system was designed as follows.

The proboscidean samples were collected and identifiedby morphology by one author (J.S.) and revealed to a‘‘witness’’ (William S. Moore of Wayne State University)but not to the remaining co-authors (H.Y. and E.M.G.)who performed the molecular test. The samples werebrought to the laboratory with only sample numbersand were tested by two authors (H.Y. and E.M.G.) whohad no prior knowledge about the age or the taxonomicstatus of these samples beyond knowing that they allbelong to the order Proboscidea. The identity of thesamples was not revealed until their nucleotide se-quences were obtained and an initial analysis of thesesequences was completed.The laboratory experiment was carried out in a plant

molecular biology laboratory where no mammalianDNA (except for human DNA) was previously handled.A stringent laboratory control procedure was applied toprevent contaminations from various sources. A set ofequipment and reagents designated solely for ancientDNA work was used. Disposable small equipment wasemployed and if reusable glassware had to be used, itwas soaked in 0.5% sodium hypochlorite for 2 h toovernight. Extraction buffers were exposed to UV lightin a UV crosslinker prior to DNA isolation. Extractionand PCR-negative controls (reagents without tissue)were used in parallel with samples on all occasions.Between 0.15 and 0.5 g dust samples were collected bydrilling into the material using a hand drill with a

disposable 3-mm bit. After EDTA treatment, sampleswere divided equally for the following three DNAextraction approaches: (a) proteinase K method (Coo-per, 1994; Hagelberg, 1994), (b) glass bead approach(Cano and Poinar, 1993), and (c) 2% CTAB-basedmethod modified from isolation protocols for planttissues (Doyle and Doyle, 1987; Golenberg, 1994b). A5-µL aliquot of the extraction result was examined in a0.7% agarose minigel containing 0.6% chemically modi-fied galactomannan with ethidium bromide staining.PCR amplification, targeting a segment of the cyto-chrome b gene of the mitochondrial genome (mtDNA),was attempted on all extractions using previouslypublished primers, and additional internal primerswere designed based on published proboscidean se-quence information (Irwin et al., 1991). Preparation ofPCR reagents was carried out in a laminar flow hood,and the PCR solution was treated under UV light for 45min before enzyme and DNA template were added. Thehot-start PCR was run on a heat-block-based thermalcycler with the following program: 3 min of initialdenaturation at 94°C followed by 40 cycles consisting ofdenaturation at 94°C (40 s), annealing at 50°C (40 s),and extension at 72°C (1 min). When the PCR productfrom the primary amplification was too weak for fur-ther sequencing, and optimization of PCR conditionscould not improve the efficiency, two-stage nested PCRamplification (40 cycles) was performed using internalprimers Elcytb65 (CTA CCC CAT CCA ACA TAT CAACAT GAT) and Elcytb320R (CGG TAT TTC AAG TTTCCG AGT ATA GGT). Both extraction and primaryPCR-negative controls were carried along in the nestedPCR as secondary controls. To confirm the origins of theamplified DNA and to identify the blind samples, wesequenced PCR products amplified from genomic DNAthat was obtained by different isolation techniquesfrom the same specimen. For samples from which DNAwas successfully isolated by only one method, at leasttwo PCR amplifications obtained from independentextractions were sequenced. PCR products were di-rectly sequenced by adopting a double-stranded Se-quenase-based dideoxy protocol (Barnard et al., 1994).Aligned sequences were analyzed for phylogenetic infer-ence, and the results were reported elsewhere (Yang etal., 1996).

RESULTS AND DISCUSSION

We noticed that the efficiency of the three isolationmethods varies for the same sample, and the choice ofextraction buffer may be critical to the success ofrecovering endogenous DNA from different types oftissue. Except for samples EL 6 and EL 29, all exam-ined samples were successfully amplified from extractsby at least one isolation method. Direct sequencing ofthese PCR products yielded unambiguous sequences.Identical sequence was obtained from PCR products

262 SHORT COMMUNICATION

amplified from the same numbered samples that wereeither extracted using different isolation methods oramplified on different occasions. For example, all threeisolation methods gave amplifiable DNA from EL 4(African elephant), and PCR products derived fromindependent isolation methods yielded identical se-quence. Table 1 summarizes the positions of nucleotidesubstitution within the aligned 228-bp segment of thecytochrome b gene obtained from each specimen incomparison with the published Loxodonta sequence(Irwin et al., 1991). Position 1 of our sequences corre-sponds to the position 14,841 of the human genome(Anderson et al., 1981). In addition to the hybrid (EL 3),completely aligned sequences from selected taxa ofthese proboscideans and their detailed phylogeneticanalyses are presented elsewhere (Yang et al., 1996).Two pairs of identical sequence (EL 1 and EL 5, EL 23and EL 32) were obtained whereas other sequences aresimilar but not identical to each other. The sequencefrom EL 4 differs from the published Loxodonta se-quence by only two nucleotides (positions 38 and 128),and both changes are third-position synonymous substi-tutions. Sequence from EL 3 (the hybrid with Elephasbeing the mother) is very similar to that of EL 1 and EL5 except for one third-position synonymous substitu-tion at position 104 that is shared with the Loxodontareference (Fig. 1 and Table 1). There are eight third-position substitutions between EL 2 (Mammuthus fromSiberia) and the published Loxodonta sequence, andseven third-position changes between EL 19 (Mammu-thus from Alaska) and the published Loxodonta se-quence. The sequences from EL 1 and EL 5 are identi-cal, and later they were revealed to come from the sameindividual Elephas. Similarly, identical sequences fromEL 23 and EL 32 matched the fact that they weresampled from different ribs of the same individualMammut. The nucleotide substitution patterns in thetested sequences are consistent with known mamma-

lian mtDNAmutation patterns in which third-positiontransitions are predominate. Phylogenetic analysis ofthese sequences (Yang et al., 1996) supports a groupingof all proboscidean sequences with Mammut as theearliest diverging taxon, which is consistent with thetraditional morphologically based classification (Sho-shani, 1996; Tassy, 1996).Reproducibility of ancient DNA results has been a

serious concern among researchers of ancient DNA,

TABLE 1

Sequencing Result from the Blind Testing, Summarizing Nucleotide Substitutions within 228-bp Cytochrome bSequence from Tested Proboscideans Relative to the Published Loxodonta Sequence (Irwin et al., 1991)

Sampleno. Taxon

GenBankaccession

no.

Nucleotide substitutions compared to published Loxodonta africana reference sequence

26 38 74 80 92 98 104 116 125 128 132 134 155 173 176 179 194 221 224

Reference L. africana X56285 A A T C T T T T T A T C A C T C C T GEL 1 E. maximus U23740 G C C T C C C C G T C T AEL 2 M. primigenius U23738 G C C T G G T N AEL 3 L. 3 E.Hybrid G C C T C C C G T C T AEL 4 L. africana U23741 C GEL 5 E. maximus U23740 G C C T C C C C G T C T AEL 19 M. primigenius U23739 G C C T C G CEL 23 M. americanum U23737 C C C C C A T C T C AEL 32 M. americanum U23737 C C C C C A T C T C A

Note. The numbers refer to the positions along the sequence of Loxodonta. A blank space represents a base identical to the publishedsequence; N indicates an unresolved base. Complete sequences are available from GenBank under the access numbers presented in the table.

FIG. 1. Autoradiograph of a segment of the cytochrome b genefrom EL 1 (Asian elephant, right) and EL 3 (a putative hybridbetween the two extant elephant genera, left). The arrow points tothe single nucleotide substitution in the hybrid EL 3 at position 104(C to T transition).

263SHORT COMMUNICATION

and the reliability of ancient DNAsequence is of criticalimportance for further phylogenetic analysis. There areseveral advantages in using a blind testing systemsimilar to the one described in this paper. This isespecially true in a study of a taxonomic lineageincluding both extinct and extant taxa from whichsequence data have been available. First, a blindtesting can serve not only as a self-checking system forlaboratory techniques but also an experimental designthat can be used for reproducing ancient DNA results;thus, such a testing is capable of establishing theauthenticity of the obtained ancient sequences. Whenidentical sequences are consistently obtained from dif-ferent samples, two possibilities may exist: (1) If thetwo samples are shown to be collected from the sameanimal or the same taxon, the identical sequences arereproducible evidence. (2) If, however, there are nomultiple samples from the same individual, or fromconspecific samples collected from close geographiclocations with contemporary age, then laboratory cross-contamination cannot be ruled out. In this study, weobtained identical sequences from EL 1 (skin) and EL 5(muscle) which were derived from a single animal(Elephas); similarly, we obtained two other identicalsequences from EL 23 and EL 32 from different ribs ofthe sameMammut skeleton. This evidence, along withthe genetic differences found in comparison with othersamples, increased our confidence in accepting them asendogenous sequences.Second, in a blind test, the similarities and differ-

ences in the sequences derived from the tested samplesin comparison with published information should begenerally consistent with the independently deter-mined identity based on morphological characters. Forexample, the sequence similarity between our EL 4 andthe published Loxodonta sequence indicated that EL 4was an African elephant, and this was consistent withthe prior identification of this animal. Considering theintraspecific variation of the cytochrome b gene and thegenetic heterogeneity among different subpopulationsof African elephants (Georgiadis et al., 1994), twothird-position synonymous substitutions within 228 bpbetween the two individuals are not surprising. Whencomparing EL 3 and EL 1 (or EL 5), the similaritysuggests a close taxonomic affinity between the twoindividuals. Indeed, EL 1 and EL 5 are derived from thesame Asian elephant, and sequence EL 3 is from aputative intrageneric hybrid between a female Asianelephant and a male African elephant. Although thepaternity of the hybrid is not established on moleculargrounds, the birth mother of this hybrid was an Asianelephant. Considering maternal inheritance of mtDNAin animals, such similarity between the cytochrome bsequences of the hybrid and an Asian elephant isexpected. In addition, the two sequences are not identi-cal, indicating no cross-contamination between the twosamples.

Third, although a double blind testing in two sepa-rate laboratories is ideal, a blind test such as the onedescribed in this paper can be satisfactorily performedin one laboratory. Because techniques used in the samelaboratory are consistent, this will reduce the chance ofa false negative result introduced by inconsistent tech-nical practice across different laboratories. Still, underthe same technical system, the characteristics of testedDNA (e.g., extraction yield and amplification efficiency)can be directly compared. For example, less DNA wasextracted from sample EL 2 (46,000-year-old woollymammoth from Siberia) and bone specimens (e.g., EL19, woolly mammoth fromAlaska) compared with DNAyield from other relatively younger aged samples. Thesecharacteristics are consistent with the difficulties ofamplifying longer fragments from these extracts, point-ing out the poor preservational status of the templateDNA in these specimens. The difference in nucleotidesequences from the tested samples rules out the possi-bility of laboratory-specific contamination or cross-contamination.Finally, a blind test also reduces bias during se-

quence analysis and phylogenetic inference. If thetaxonomic status of tested taxa is known prior tophylogenetic analysis, selection of characters or accep-tance (or rejection) of inferred trees may be biasedtoward the expected result. Therefore, phylogeneticinference based on sequences that were derived from ablind testing enhances the independence of such analy-ses. The fact that our sequence-based phylogeny (Yanget al., 1996) matches the traditional morphologicallybased classification of proboscideans (Shoshani, 1996;Tassy, 1996) not only provides an independent test forthe traditional proboscidean taxonomy but also servesas further testimony for the authenticity of these DNAsequences.It should be pointed out, however, that such a blind

testing system may have limited power for populationlevel investigations in which multiple individuals mayyield identical sequence. In other cases while examin-ing a variety of unrelated taxa using different sets ofprimers, such blind testingmay not be applicable. If thegeneticist selects the sample and the specimens haverecognizable colors during extraction, then a blinddesign may be used at PCR and sequencing stages totest the consistency of ancient DNA results.

ACKNOWLEDGMENTS

This work was supported by an Alfred P. Sloan PostdoctoralFellowship (93-4-6-ME) awarded to HongYang.We thankWilliams S.Moore (Wayne State University) for being the observer in the blindtesting and for providing primers during initial experiments. We arealso grateful to Richard E. Tashian (University of Michigan) andthree anonymous reviewers for valuable suggestions.

264 SHORT COMMUNICATION

REFERENCES

Anderson, S., Bankier, A. T., Barrell, B. G., de Bruijn, M. H. L.,Coulson, A. R., Drouin, J., Eperon, I. C., Nierlich, D. P., Roe, B. A.,Sanger, F., Schreier, P. H., Smith, A. J. H., Staden, R., and Young,I. G. (1981). Sequence and organization of the human mitochon-drial genome.Nature 290: 457–465.

Barnard, G. F., Puder, M., Begum, N. A., and Chen, L. B. (1994). PCRproduct sequencing with [a-33P] and [a-32P]dATP. BioTechniques16: 572–573.

Beraud-Colomb, E., Roubin, R., Martin, J., Maroc, N., Gardeisen, A.,Trabuchet, G., and Goossens, M. (1995). Human b-globin genepolymorphisms characterized in DNAextracted from ancient bones12,000 years old. Am. J. Hum. Genet. 57: 1267–1274.

Cano, R. J., and Poinar, H. N. (1993). Rapid isolation of DNA fromfossil and museum specimens suitable for PCR. BioTechniques 15:432–435.

Cooper, A. (1994). DNA from museum specimens. In ‘‘Ancient DNA’’(B. Herrmann and S. Hummel, eds.), pp. 149–165, Springer-Verlag,NewYork.

Doyle, J. J., and Dolye, J. L. (1987). A rapid DNA isolation procedurefor small quantities of fresh leaf tissue. Phytochem. Bull. 19: 11–15.

Georgiadis, N., Bischof, J., Templeton, A., Patton, J., Karesh, W., andWestern, D. (1994). Structure and history of Africanelephant populations. I. Eastern and southernAfrica. J. Hered. 85:100–104.

Golenberg, E. M. (1994a). Uncertainty in the analysis of fossil DNAsequences. In ‘‘Conservation of Plant Genes II: Utilization ofAncient and Modern DNA’’ (R. P. Adams, R. J. Miller, E. M.Golenberg, and J. E. Adams, Eds.), pp. 367–692, Missouri Botani-cal Garden.

Golenberg, E. M. (1994b). Fossil samples: DNA from plant compres-sion fossils. In ‘‘Ancient DNA’’ (B. Herrmann and S. Hummel, Eds.),pp. 233–252, Springer Verlag, NewYork.

Hagelberg, E. (1994). Mitochondrial DNA from ancient bones. In‘‘Ancient DNA’’ (B. Herrmann and S. Hummel, Eds.), pp. 195–204,Springer-Verlag, NewYork.

Handt, O., Hoss, M., Krings, M., and Paabo, S. (1994a).Ancient DNA:Methodological challenges. Experientia 50: 524–529.

Handt, O., Richards, M., Trommsdorff, M., Kilger, C., Simanainen, J.,Georgiev, O., Bauer, K., Stone, A., Hedges, R., Schaffner, W.,Utermann, G., Sykes, B., and Paabo, S. (1994b). Molecular geneticanalyses of the Tyrolean ice man. Science 264: 1775–1778.

Hanni, C., Laudet, V., Coll, J., and Stehelin, D. (1994a). An unusualmitochondrial DNA sequence variant from an Egyptian mummy.Genomics 22: 487–489.

Hanni, C., Laudet, V., Stehelin, D., and Taberlet, P. (1994b). Trackingthe origins of the cave bear (Ursus spelaeus) by mitochondrial DNAsequencing. Proc. Natl. Acad. Sci. USA 91: 12336–12340.

Hoss, M., Dilling, A., Currant, A., and Paabo, S. (1996). Molecularphylogeny of the extinct ground sloth Mylodon darwinii. Proc.Natl. Acad. Sci. USA 93: 181–185.

Irwin, D. M., Kocher, T. D., and Wilson, A. C. (1991). Evolution of thecytochrome b gene of mammals. J. Mol. Evol. 32: 128–144.

Lindahl, T. (1993). Recovery of antediluvian DNA.Nature 365: 700.Lowenstein, J. M., and Shoshani, J. (1996). Proboscidean relation-ships based on immunological data. In ‘‘The Proboscidea: Evolutionand Palaeoecology of Elephants and Their Relatives’’ (J. Shoshaniand P. Tassy, Eds.), pp. 49–54, Oxford Univ. Press, Oxford.

Richards, M. B., Sykes, B. C., and Hedges, R. E. M. (1995). Authenti-cating DNA extracted from ancient skeletal remains. J. Arch. Sci.22: 291–299.

Shoshani, J. et al. (76 co-authors) (1982). On the dissection of afemale Asian elephant (Elephas maximus maximus Linnaeus,1758) and data from other elephants. Elephant 2: 3–93.

Shoshani, J., Lowenstein, J. M.,Walz, D.A., and Goodman,M. (1985).Proboscidean origins of mastodon and woolly mammoth demon-strated immunologically. Paleobiology 11: 429–437.

Shoshani, J., Fisher, D. C., Zawiskie, J. M., Thurlow, S. J., Shoshani,S. L., Benninghoff, W. S., and Zoch, F. H. (1989). The Sheltonmastodon site: Multidisciplinary study of a late Pleistocene (Two-creekan) locality in southeastern Michigan. Contrib. Mus. Paleon-tol. Univ. Mich. 27: 393–436.

Shoshani, J. (1996). Para- or monophyly of the gomphotheres andtheir position within Proboscidea. In ‘‘The Proboscidea: Evolutionand Palaeoecology of Elephants and Their Relatives’’ (J. Shoshaniand P. Tassy, Eds.), pp. 149–177, Oxford Univ. Press, Oxford.

Taylor, P. G. (1996). Reproducibility of ancient DNA sequences fromextinct Pleistocene fauna.Mol. Bio. Evol. 13: 283–285.

Tassy, P. (1996). Who is who among the Proboscidea. In ‘‘TheProboscidea: Evolution and Palaeoecology of Elephants and TheirRelatives’’ (J. Shoshani and P. Tassy, Eds.), pp. 39–48, Oxford Univ.Press, Oxford.

Yang, H., Golenberg, E. M., and Shoshani, J. (1996). Phylogeneticresolution within Elephantidae using fossil DNA sequence fromAmerican mastodon (Mammut americanum) as an outgroup. Proc.Natl. Acad. Sci. USA 93: 1190–1194.

265SHORT COMMUNICATION