Cell, Vol. 90, 19-30, July 11, 1997, Copyright @1997 by Cell Press Neandertal DNA Sequences and the origin of Modern Humans Matthias Krings, * Anne Stone, t Ralf W. Schmitz, t these analyses rely on assumptions, such as the ab Heike Krainitzki, 5 Mark Stoneking, t and Svante Paabo* sence of selection and a clock-like rate of mole volution in the DNA sequences under study, whose University of Munich alidity has been questioned (Wolpoff, 1989: Templeton Box202136 992). An additional and more direct way to address the D-80021 Munich question of the relationship between modern humans nd Neandertals would be to analyze DNA sequences t Department of Anthropology from the remains of neandertals nsylvania State University The reproducible retrieval of ancient DNA sequences State College, Pennsylvania 16802 became possible with the invention of the polymerase t Rheinisches Amt fur Bodendenkmalpflege hain reaction (Mullis and Faloona, 1987: Paabo et al Endenicher Strasse 133 D-53115 Bonn ilson, 1991; Lindahl 1993a) as well as empirical studies Germany (Paabo, 1989: Hoss et al. 1996a), show that dNa in 5Hohere berufsfachschule fur fossil remains is highly affected by hydrolytic as well praparationstechnische Assistenten as oxidative damage. Therefore, the retrieval of DNA Markstrasse 185 sequences older than about 100,000 years is expected 44799 Bochum to be difficult, if not impossible, to achieve(Paabo and Germany Wilson, 1991). Fortunately, Neandertal remains fall most fail to yield ampl的hym within the age range that in principle allow uences to survive. It is noteworth at even Summary mong remains that are yo DNA was extracted from the Neandertal-type speci- 996b) In addition, contamination of ancient specimens men found in 1856 in western Germany By sequencing and extracts with modern DNA poses a serious probler clones from short overlapping PCR products, a hith (Handt et al. 1994a)thatrequires numerous precautions erto unknown mitochondrial (mt) DNA sequence was ind controls. This is particularly the case when human determined. Multiple controls indicate that this se- remains are studied, since human dna is the most com quence is endogenous to the fossil. Sequence com- mon source of contamination. Therefore, a number of arisons with human mtDNA sequences, as well as tic analyses, show that the Neandertal se- mined from extracts of an ancient specimen can be quence falls outside the variation of modern humans. taken to be genuine (Paabo et al. 1989 Lindahl, 1993b Furthermore, the age of the common ancestor of the Handt et al. 1994a: Handt et al. 1996 Neandertal and modern human mtDNAs is estimated Since 1991, the Neandertal-type specimen, found in to be four times greater than that of the common an- 1856 near Dusseldorf, Germany, has been the subject cestor of human mt DNAS. This suggests that Neander- of an interdisciplinary project of the Rheinisches tals went extinct without contributing mtDNA to mod Landesmuseum Bonn, initiated and led by R. W. s enn numan (Schmitz et al. 1995: Schmitz, 1996). As a part of thi project, a sample was removed from the Neandert specimen for DNA analysis. Here, we present the se- quence of a hypervariable part of the mtDNA control region derived from this sample. We describe the evi Neandertals are a group of extinct hominids that inhab nce in support of its authenticity and analyze the rela ited Europe and western Asia from about 300,000 to ionship of this sequence to the contemporary human 30,000 years ago. During part of this time they coexisted mtDNa gene pool with modern humans. Based on morphological compari sons, it has been variously claimed that Neandertals: (1)were the direct ancestors of modern Europeans; (2) Results contributed some genes to modern humans; or(3)were completely replaced by modern humans without con- Amino Acid racemization tributing any genes (reviewed in Stringer and Gamble, 3.5 g section of the right humerus was removed from 1993: Trinkaus and Shipman 1993: Brauer and Stringer, the Neandertal fossil(Figure 1). It has previously been 1997). Analyses of molecular genetic variation in the shown that ancient specimens exhibiting high levels of mitochondrial and nuclear genomes of contemporary amino acid racemization do not contain sufficient DNA human populations have generally supported the third or analysis(Poinar et al., 1996). To investigate wheti view, i.e., that Neandertals were a separate species that the state of preservation of the fossil is compatible with went extinct without contributing genes to modern hu- DNA retrieval, we therefore analyzed the extent of amino mans(Cann et al. 1987; Vigilant et al., 1991; Hammer, acid racemization. Samples of 10 mg were removed 1995; Armour et al., 1996 Tishkoff et al., 1996). However, from the periostal surface of the bone from the compact
Cell, Vol. 90, 19–30, July 11, 1997, Copyright 1997 by Cell Press Neandertal DNA Sequences and the Origin of Modern Humans Matthias Krings,* Anne Stone,† Ralf W. Schmitz,‡ these analyses rely on assumptions, such as the abHeike Krainitzki, sence of selection and a clock-like rate of molecular § Mark Stoneking,† and Svante Pa¨ a¨ bo* *Zoological Institute evolution in the DNA sequences under study, whose University of Munich validity has been questioned (Wolpoff, 1989; Templeton, PO Box 202136 1992). An additional and more direct way to address the D-80021 Munich question of the relationship between modern humans Germany and Neandertals would be to analyze DNA sequences from the remains of Neandertals. †Department of Anthropology Pennsylvania State University The reproducible retrieval of ancient DNA sequences State College, Pennsylvania 16802 became possible with the invention of the polymerase ‡Rheinisches Amt fu¨r Bodendenkmalpflege chain reaction (Mullis and Faloona, 1987; Pa¨ a¨bo et al., Endenicher Strasse 133 1989). However, theoretical considerations, (Pa¨ a¨bo and D-53115 Bonn Wilson, 1991; Lindahl 1993a) as well as empirical studies Germany (Pa¨ a¨bo, 1989; Ho¨ ss et al., 1996a), show that DNA in §Ho¨ here Berufsfachschule fu¨r fossil remains is highly affected by hydrolytic as well pra¨parationstechnische Assistenten as oxidative damage. Therefore, the retrieval of DNA Markstrasse 185 sequences older than about 100,000 years is expected D-44799 Bochum to be difficult, if not impossible, to achieve (Pa¨ a¨bo and Germany Wilson, 1991). Fortunately, Neandertal remains fall within the age range that in principle allows DNA sequences to survive. It is noteworthy, though, that even Summary among remains that are younger than 100,000 years most failto yield amplifiable DNA sequences(Ho¨ ss et al., DNA was extracted from the Neandertal-type speci- 1996b). In addition, contamination of ancient specimens men found in 1856 in western Germany. By sequencing and extracts with modern DNA poses a serious problem clones from short overlapping PCR products, a hith- (Handt et al., 1994a) that requires numerous precautions and controls. This is particularly the case when human erto unknown mitochondrial (mt) DNA sequence was determined. Multiple controls indicate that this se- remains are studied, since human DNA is the most comquence is endogenous to the fossil. Sequence com- mon source of contamination. Therefore, a number of parisons with human mtDNA sequences, as well as criteria need to befulfilled before a DNA sequence deterphylogenetic analyses, show that the Neandertal se- mined from extracts of an ancient specimen can be quence falls outside the variation of modern humans. taken to be genuine (Pa¨a¨bo et al., 1989; Lindahl, 1993b; Furthermore, the age of the common ancestor of the Handt et al., 1994a; Handt et al., 1996). Neandertal and modern human mtDNAs is estimated Since 1991, the Neandertal-type specimen, found in 1856 near Du¨ sseldorf, Germany, has been the subject to be four times greater than that of the common ancestor of human mtDNAs. This suggests that Neander- of an interdisciplinary project of the Rheinisches Landesmuseum Bonn, initiated and led by R. W. S. tals went extinct without contributing mtDNA to modern humans. (Schmitz et al., 1995; Schmitz, 1996). As a part of this project, a sample was removed from the Neandertal specimen for DNA analysis. Here, we present the sequence of a hypervariable part of the mtDNA control Introduction region derived from this sample. We describe the evidence in support of its authenticity and analyze the rela- Neandertals are a group of extinct hominids that inhab- tionship of this sequence to the contemporary human ited Europe and western Asia from about 300,000 to mtDNA gene pool. 30,000 years ago. During part of this time they coexisted with modern humans. Based onmorphological comparisons, it has been variously claimed that Neandertals: (1) were the direct ancestors of modern Europeans; (2) Results contributed some genes to modern humans; or (3) were completely replaced by modern humans without con- Amino Acid Racemization tributing any genes (reviewed in Stringer and Gamble, A 3.5 g section of the right humerus was removed from 1993; Trinkaus and Shipman 1993; Bra¨ uer and Stringer, the Neandertal fossil (Figure 1). It has previously been 1997). Analyses of molecular genetic variation in the shown that ancient specimens exhibiting high levels of mitochondrial and nuclear genomes of contemporary amino acid racemization do not contain sufficient DNA human populations have generally supported the third for analysis (Poinar et al., 1996). To investigate whether view, i.e., that Neandertals were a separate species that the state of preservation of the fossil is compatible with went extinct without contributing genes to modern hu- DNA retrieval, we therefore analyzed the extent of amino mans (Cann et al., 1987; Vigilant et al., 1991; Hammer, acid racemization. Samples of 10 mg were removed 1995; Armour et al., 1996; Tishkoff et al., 1996). However, from the periostal surface of the bone, from the compact
tends to be degraded and damaged to an extent that makes amplification of segments of mtDNA longer than 100-200 bp difficult(Paabo, 1989). Therefore, two prim ers(L16, 209, H16, 271) that amplify a 105-bp-segment of the human mtDNA control region(including primers were used to perform amplifications from the bone ex- tract as well as from an extraction control. a tion product was obtained in the bone extract but not in the control(data not shown). In a subsequent experi- ment, this was repeated and the same results were ob Sequence Variation of the Amplification Product igure 1. Sample Removed from the Right Humerus of the Neander- The two amplification products were cloned in a plasmid tal-Type Specimen vector and 18 and 12 clones, respectively were se- uenced(Figure 2, extract A). Twenty-two of the 30 cortical bone and from the endostal surface of the mar clones contained seven nucleotide substitutions and row cavity. Samples were also removed from remnants one insertion of an adenine residue, when compared to of a varnish, with which the specimen has been treated the standard human reference sequence(Anderson et at least twice. The samples were hydrolyzed under acid al., 1981). Three of these eight differences to the refer conditions, and the released amino acids were analy. ence sequence were individually lacking in a total of five using high performance liquid chromatography and fluo- of the clones. In addition, among the 27 clones were rescent detection(Poinar et al., 1996). Table 1 shows nine differences that each occurred in one clone, three that the total amounts of the amino acids detected in differences that occurred in two clones and one that the neandertal bone are 20%-73% of those in modern occurred in three clones, respectively. Such changes bone and more than two orders of magnitude higher that are present in only a few clones are likely to be than in the varnish, indicating that the results do not due to misincorporations by the DNA polymerase during more, the absolute and relative amounts of the amino DNA. In addition some of these could be due to mito acids analyzed (e.g, the ratio of glycine to aspartic acid chondrial heteroplasmy, which may be more common are similar in the three Neandertal samples and compa in humans than often assumed(Comas et al. 1995 rable to those of a contemporary bone. Most impor Ivanovet al 1996)and is abundant in some mammalian tantly, the ratio of the D to the L enantiomers of aspartic species(Petriet al 1996) Of theremaining three clones, acid in the three Neandertal samples varies between two were identical to the reference sequence, and the 0 11 and 0. 12, which is in the range compatible with third clone differed from the reference sequence at one DNA survival( Poinar et al. 1996). Thus, the extent of position amino acid racemization in the Neandertal fossil sug Thus, the amplification product was composed of two gests that it may contain amplifiable DNA three clones that is similar to the human reference se- DNA Extraction and Amplification quence, and another class represented by 27 clones DNA was extracted from 0. 4 g of the cortical compact hat exhibits substantial differences from it. the former bone. Previous experience shows that ancient dna class of molecules probably reflects contamination of Table 1. Racemization Results for Three Neandertal Bone Samples, Varnish from the Neandertal Fossil, and Modern Bone Periostal Surface Compact Bone Endostal surface Modern bone Valine(%) 23 0.11 0.114 0.110 D/L alanine 0.006 0.007 0.004 008 D/L leucine amino acid analysis of the Neandertal bone, va emoved from the bone surface, and a two-year-old bone sample no acid compositions in percentages of the eight yzed, and the D/L-ratios for three amino acids. NI electable d form
Cell 20 tends to be degraded and damaged to an extent that makes amplification of segments of mtDNA longer than 100–200 bp difficult (Pa¨a¨ bo, 1989). Therefore, two primers (L16,209, H16,271) that amplify a 105-bp-segment of the human mtDNA control region (including primers) were used to perform amplifications from the bone extract as well as from an extraction control. An amplification product was obtained in the bone extract but not in the control (data not shown). In a subsequent experiment, this was repeated and the same results were obtained. Sequence Variation of the Amplification Product Figure 1. Sample Removed from the Right Humerus of the Neander- The two amplification products were cloned in a plasmid tal-Type Specimen vector and 18 and 12 clones, respectively, were sequenced (Figure 2, extract A). Twenty-two of the 30 cortical bone, and from the endostal surface of the mar- clones contained seven nucleotide substitutions and row cavity. Samples were also removed from remnants one insertion of an adenine residue, when compared to of a varnish, with which the specimen has been treated the standard human reference sequence (Anderson et al., 1981). Three of these eight differences to the refer- at least twice. The samples were hydrolyzed under acid ence sequence were individually lacking in a total of five conditions, and the released amino acids were analyzed using high performance liquid chromatography and fluo- of the clones. In addition, among the 27 clones were rescent detection (Poinar et al., 1996). Table 1 shows nine differences that each occurred in one clone, three differences that occurred in two clones and one that that the total amounts of the amino acids detected in the Neandertal bone are 20%–73% of those in modern occurred in three clones, respectively. Such changes bone and more than two orders of magnitude higher that are present in only a few clones are likely to be than in the varnish, indicating that the results do not due to misincorporations by the DNA polymerase during reflect the amino acid content of the varnish. Further- PCR, possibly compounded by damage in the template more, the absolute and relative amounts of the amino DNA. In addition, some of these could be due to mitoacids analyzed (e.g., the ratio of glycine to aspartic acid) chondrial heteroplasmy, which may be more common are similar in the three Neandertal samples and compa- in humans than often assumed (Comas et al., 1995; rable to those of a contemporary bone. Most impor- Ivanov et al., 1996) and is abundant in some mammalian tantly, the ratio of the D to the L enantiomers of aspartic species (Petri et al., 1996). Of theremaining three clones, acid in the three Neandertal samples varies between two were identical to the reference sequence, and the 0.11 and 0.12, which is in the range compatible with third clone differed from the reference sequence at one DNA survival (Poinar et al., 1996). Thus, the extent of position. amino acid racemization in the Neandertal fossil sug- Thus, the amplification product was composed of two gests that it may contain amplifiable DNA. classes of sequences, a minor class represented by three clones that is similar to the human reference seDNA Extraction and Amplification quence, and another class represented by 27 clones DNA was extracted from 0.4 g of the cortical compact that exhibits substantial differences from it. The former bone. Previous experience shows that ancient DNA class of molecules probably reflects contamination of Table 1. Racemization Results for Three Neandertal Bone Samples, Varnish from the Neandertal Fossil, and Modern Bone Periostal Surface Compact Bone Endostal Surface Varnish Modern Bone Total (ppm) 23,167 83,135 53,888 145 113,931 Aspartic acid (%) 7.8 8.3 7.4 10 8.3 Serine (%) 0.7 0.7 0.7 2 0.6 Glutamic acid (%) 20.2 20.1 20.2 22 19.9 Glycine (%) 49.5 49.0 50.2 22 51.8 Alanine (%) 14.4 14.0 14.0 11 11.1 Valine (%) 3.5 3.9 3.9 23 3.9 Isoleucine (%) 0.5 0.5 0.6 1 0.7 Leucine (%) 3.4 3.3 3.2 9 3.6 Glycine/aspartic acid 6.3 5.9 6.8 2.1 6.2 D/L aspartic acid 0.117 0.114 0.110 ND 0.05 D/L alanine 0.006 0.007 0.004 0.08 0.01 D/L leucine 0.005 ND ND ND ND Comparison of the amino acid analysis of the Neandertal bone, varnish removed from the bone surface, and a two-year-old bone sample. Given are the total amounts of the amino acids analyzed (ppm, parts per million), the amino acid compositions in percentages of the eight amino acids analyzed, and the D/L-ratios for three amino acids. ND, no detectable D form
CAGCAATCRACCCTCAACTATCACACATCAACTOCAACTCcRAAGCCACCCCT-CRCCC c5001002040 Figure 3. Quantitation of the Putative Neandertal mtDNA with a 12 bp of cd LWcompetitor molecules added are indicated. The control R17: amplification (C)contained neither competitor nor Neandertal ex- 1996). Therefore, the number of template molecules rep resenting the putative Neandertal sequence in the ex tract was determined by quantitative PCR. To this end a molecule representing the putative Neandertal se- quence but carrying a 12 bp deletion was constructed To each step in a dilution series of this construct, a constant amount of extract was added and amplifica tions were performed using primers that are specific for a 104 bp product of the putative Neandertal sequenc Figure 2. The DNA Sequ of Clones Derived from Four Amplifi and that do not amplify contemporary human se- cations of the Mitochondrial Control Region from the Neandertal uences. The results(Figure 3)show that on the order of 10 putative Neandertal molecules exist per microliter man reference sequence(Anderson e of extract and thus that amplifications starting from 5 al, 1981)given above. The clone designations consist of a letter(A, I of extract are initiated from approximately 50 template B, C)indicating the DNA extract followed by a number indicating ion, as well as a number after the period molecules. however due to variation in the efficien dentifying the particular clone. Extracts A and B were performed at the University of Munich: extract C, at Penn State University number of template molecules added to an individual Clones derived from different amplifications are separated by a amplification, some amplifications may start from fewer (or even single)molecules. This makes nucleotide misin- one clone differs from the majority of sequence corporations in earby cycles of the amplification reaction pper amplifications (performed at the University of Munich) primers 16, 209(5'-CCC CAT GCT TAC AAG CAA GT-3)and H16, 271(5 ikely to affect a large proportion of the molecules in the ers NL16, 230(5-GCA CAG CAA TCA ACC TTC AAC TG-3) and carry miscoding base modifications(Hoss et al. 1996a) NH16, 262(5'-GTA GAT TTG TTG ATA TCC TAG TGG GTG TAA-3) To detect this type of sequence change, amplifications were used were performed such that each sequence position to be determined was covered by at least two independent PCR reactions. The products of each PCR reaction were the specimen, which is likely to have occurred during independently cloned and the sequences determined handling and treatment of the specimen during the 140 from multiple clones years since its discovery. The other class of sequences is not obviously of modern origin. Further experiments Authenticity of Sequences were therefore performed to determine if this class is The inadvertent amplification of small amounts of con- endogenous to the Neandertal fossil temporary DNA is a major source of erroneous results in the study of ancient DNA sequences(Paabo et al 1989: Lindahl, 1993b; Handt et al. 1994a. Such contami- Quantitation of putative Neandertal dna lation may result in the amplification of not only contem Amplifications that start from more than 1000 ancient porary organellar mtDNA template molecules tend to yield reproducible results of mtDNA(Collura and Stewart, 1995; van der Kuyl et while amplifications starting from fewer molecules tend al., 1995; Zischler et al., 1995). Several experiments were to yield results that vary between experiments, due to performed in order to exclude modern DNA, including a misincorporations during the early cycles of the PCR as nuclear insertion of mtDNA, as the source of the putative ell as due to sporadic contamination(Handt et al
Neandertal mtDNA Sequence 21 Figure 3. Quantitation of the Putative Neandertal mtDNA A dilution series of a competitor construct carrying the putative Neandertal sequence with a 12 bp deletion was added to 2.5 ml of extract A from the fossil. Primers used were specific for the putative Neandertal sequence. Above the lanes, the approximate numbers of cd LWcompetitor molecules added are indicated. The control amplification (C) contained neither competitor nor Neandertal extract. 1996). Therefore, the number of template molecules representing the putative Neandertal sequence in the extract was determined by quantitative PCR. To this end, a molecule representing the putative Neandertal sequence but carrying a 12 bp deletion was constructed. To each step in a dilution series of this construct, a constant amount of extract was added and amplifications were performed using primers that are specific for a 104 bp product of the putative Neandertal sequence Figure 2. The DNA Sequences of Clones Derived from Four Amplifi- and that do not amplify contemporary human secations of the Mitochondrial Control Region from the Neandertal quences. The results (Figure 3) show that on the order Fossil of 10 putative Neandertal molecules exist per microliter Dots indicate identity to a human reference sequence (Anderson et of extract and thus that amplifications starting from 5 al., 1981) given above. The clone designations consist of a letter (A, ml of extract are initiated from approximately 50 template B, C) indicating the DNA extract followed by a number indicating molecules. However, due to variation in the efficiency the amplification reaction, as well as a number after the period of individual primer pairs, and stochastic variation in the identifying the particular clone. Extracts A and B were performed number of template molecules added to an individual at the University of Munich; extract C, at Penn State University. Clones derived from different amplifications are separated by a amplification, some amplifications may start from fewer blank line. Asterisks identify sequence positions where more than (or even single) molecules. This makes nucleotide misinone clone differs from the majority of sequences. For the three corporations in early cycles of the amplification reaction upper amplifications (performed at the University of Munich) primers likely to affect a large proportion of the molecules in the L16,209 (59-CCC CAT GCT TAC AAG CAA GT-39) and H16,271 (59- final amplification product. Such misincorporations may GTG GGT AGG TTT GTT GGT ATC CTA-39) were used. For the bottom amplification (performed at Penn State University) the prim- be frequent since the template molecules are likely to ers NL16,230 (59-GCA CAG CAA TCA ACC TTC AAC TG-39) and carry miscoding base modifications (Ho¨ ss et al., 1996a). NH16,262 (59-GTA GAT TTG TTG ATA TCC TAG TGG GTG TAA-39) To detect this type of sequence change, amplifications were used. were performed such that each sequence position to be determined was covered by at least two independent PCR reactions. The products of each PCR reaction were the specimen, which is likely to have occurred during independently cloned and the sequences determined handling and treatment of the specimen during the 140 from multiple clones. years since its discovery. The other class of sequences is not obviously of modern origin. Further experiments Authenticity of Sequences were therefore performed to determine if this class is The inadvertent amplification of small amounts of conendogenous to the Neandertal fossil. temporary DNA is a major source of erroneous results in the study of ancient DNA sequences (Pa¨ a¨bo et al., 1989; Lindahl, 1993b; Handt et al.,1994a). Such contamiQuantitation of Putative Neandertal DNA nation may result inthe amplification of not only contemAmplifications that start from more than 1000 ancient porary organellar mtDNA but also of nuclear insertions template molecules tend to yield reproducible results, of mtDNA (Collura and Stewart, 1995; van der Kuyl et while amplifications starting from fewer molecules tend al., 1995; Zischler et al., 1995). Several experiments were to yield results that vary between experiments, due to performed in order to exclude modern DNA, including a misincorporations during the early cycles of the PCR as nuclear insertion of mtDNA,as the source of the putative well as due to sporadic contamination (Handt et al., Neandertal sequence
Since nuclear insertions are less numerous than mito from 0. 4 g of the bone. When the primers L16, 209 and hondrial genomes in the organelles, any single insertion H16, 271 were used in an amplification from this extract sequence is expected to represent a major proportion and the product cloned( Figure 2, extract B), ten clones of an amplification product only in cases where a prime arried the eight differences from the reference se- favors the amplification of an insertion sequence over quence observed in the amplifications from the first ex the corresponding mtDNA sequence. This occurs when tract, as well as two changes affecting single clones. mismatches to the primer in the mtDNA make the prim- In addition, three sequence positions carried changes ing of an insertion more efficient than that of the organel occurring in five and four clones. These changes were lar mtDNA(Handt et al., 1996). Therefore, the preferential not observed in combination in the previous four amplifi amplification of an insertion sequence is expected to cations covering this sequence segment. Since they oc- be restricted to a particular primer. In order to elucidate curred in only one amplification product, they are proba whether the putative Neandertal sequence is seen only bly due to polymerase errors in the early cycles of the when a particular primer is used, primers were ex- PCR, possibly compounded by in vitro recombination changed such that first the 5 primer was replaced by induced by damage and degradation of template DNA a primer located outside the previous amplification prod- molecules(Paabo et al. 1990). Four clones were similar ct(L16, 122), and 13 clones of this amplification product to the human reference sequence. Thus, although the were sequenced (Figure 4, clones A7 1-13). All 13 clones amplification products clearly derive from few template showed the same eight differences from the reference molecules, the putative Neandertal sequence is present sequence that were previously observed, as well as nine in a DNA extract independently prepared from the fossil differences in the region that was not included in the To further investigate whether the results are due to earlier amplification. In addition, one difference was laboratory-specific artifacts or contamination, an addi observed in one clone, as well as length variation in tional bone sample of 0. 4 g was sent to the Anthr a homopolymer of cytosine residues, previously de- cal Genetics Laboratory at Pennsylvania State Univer- scribed to be of variable length in humans( Bendall and sity where a DNA extraction was performed. When the Sykes, 1995) primers(L16, 209 and H16, 271), which had previous The 3'-primer from the first amplification was the replaced by a primer (H16, 379)located outside the initial Neandertal sequence and contemporary human mtDNA amplification product, and 13 clones of this amplification sequences(Figure 2)were used in amplifications from product were sequenced(Figure 5, clones A121-13). this extract, 15 of the resulting clones yielded a DNA All 13 clones contained the same eight differences in sequence that was identical to the experimenter(A s ) the region overlapping the previous amplifications, as while two yielded sequences that differed by one and well as seven differences in the region not covered in two substitutions from the reference sequence, respec- the previous amplifications. In addition, two substitu- tively. However, when primers specific for the putative tions and one deletion occurred in one clone, and one Neandertal sequence(NL16, 230 and NH16, 262)were other substitution occurred in a different clone furthe used, 5 out of 5 clones yielded the putative Neandertal more, in a subsequent amplification from another ex- sequence(Figure 2, extract C). Thus, while this third tract where both primers (L16, 254-H16, 379)differed independent extract contains a larger amount of con- from the initial amplification, all four differences located temporary human DNA, probably stemming from labora in the segment included in the first amplifications were ory contamination, it confirms that the putative Nean observed in all 8 clones sequenced ( Figure 5, clones dertal sequence is present in the fossil specime B13.1-8). Thus, the retrieval of the putative Neandertal quence is not dependent on the primers used. Fu tive Neandertal sequence does not originate from a nu thermore, most primer combinations yield a large ex- clear mtDNA insertion and that it is endogenous to the cess of clones representing the putative Neandertal se fossil. it furthermore falls outside the variation of the quence over clones similar to contemporary human mtDNa gene pool of modern humans (see below). We tDNA therefore conclude that it is derived from the mitochon To further exclude the possibility that the sequence drial genome of the Neandertal individuaL. may represent a nuclear insertion, primers for the puta tive Neandertal sequence were constructed that do not amplify human mtDNA In control experiments where Determination of the Neandertal mtDNA Sequence various amounts of a cloned copy of the putative Nean. The entire sequence of hypervariable region I of the dertal sequence were mixed with human DNA, these mtDNA control region(positions 16,023 to 16, 400: An erson et al. 1981) was determine cloned sequence in 50 ng of total human DNA, i.e., less preservation of the DNA allowed only short fragments to than one copy per genome equivalent(data not shown). be amplified, this was achieved by several overlapping When these primers were used to amplify DNA isolated amplifications. Furthermore, since the quantitation ex- from 15 Africans, 6 Europeans, and 2 Asians, no amplifi eriments indicated that some al ations might start cation products were obtained(data not shown), indicat- from single molecules, and thus that misincorporations ing that this sequence is not present in the genome of in early cycles of the amplification might be misinter modern humans reted as sequence differences(Handt et al. 1996), all To test whether the extraction and amplification of sequence positions were determined from at least two the putative Neandertal sequence is reproducible, an independent amplifications At five sequence positions dditional independent DNA extraction was performed two amplifications yielded discordant results, i.e., all
Cell 22 Since nuclear insertions are less numerous than mito- from 0.4 g of the bone. When the primers L16,209 and chondrial genomes inthe organelles, any single insertion H16,271 were used in an amplification from this extract sequence is expected to represent a major proportion and the product cloned (Figure 2, extract B), ten clones of an amplification product only in cases where a primer carried the eight differences from the reference sefavors the amplification of an insertion sequence over quence observed in the amplifications from the first exthe corresponding mtDNA sequence. This occurs when tract, as well as two changes affecting single clones. mismatches to the primer in the mtDNA make the prim- In addition, three sequence positions carried changes ing of an insertion more efficient than that of the organel- occurring in five and four clones. These changes were lar mtDNA (Handt et al., 1996). Therefore, thepreferential not observed in combination in the previous four amplifiamplification of an insertion sequence is expected to cations covering this sequence segment. Since they ocbe restricted to a particular primer. In order to elucidate curred in only one amplification product, they are probawhether the putative Neandertal sequence is seen only bly due to polymerase errors in the early cycles of the when a particular primer is used, primers were ex- PCR, possibly compounded by in vitro recombination changed such that first the 59 primer was replaced by induced by damage and degradation of template DNA a primer located outside the previous amplification prod- molecules (Pa¨a¨bo et al., 1990). Four clones were similar uct (L16,122), and 13 clones of this amplification product to the human reference sequence. Thus, although the were sequenced (Figure 4, clones A7.1–13). All 13 clones amplification products clearly derive from few template showed the same eight differences from the reference molecules, the putative Neandertal sequence is present sequence that were previously observed, as well as nine in a DNA extract independently prepared from the fossil. differences in the region that was not included in the To further investigate whether the results are due to earlier amplification. In addition, one difference was laboratory-specific artifacts or contamination, an addiobserved in one clone, as well as length variation in tional bone sample of 0.4 g was sent to the Anthropologia homopolymer of cytosine residues, previously de- cal Genetics Laboratory at Pennsylvania State Universcribed to be of variable length in humans (Bendall and sity where a DNA extraction was performed. When the Sykes, 1995). primers (L16,209 and H16,271), which had previously The 39-primer from the first amplification was then resulted in a product that contained both the putative replaced by a primer (H16,379) located outside the initial Neandertal sequence and contemporary human mtDNA amplification product, and 13clones of this amplification sequences (Figure 2) were used in amplifications from product were sequenced (Figure 5, clones A12.1–13). this extract, 15 of the resulting clones yielded a DNA All 13 clones contained the same eight differences in sequence that was identical to the experimenter (A. S.), the region overlapping the previous amplifications, as while two yielded sequences that differed by one and well as seven differences in the region not covered in two substitutions from the reference sequence, respecthe previous amplifications. In addition, two substitu- tively. However, when primers specific for the putative tions and one deletion occurred in one clone, and one Neandertal sequence (NL16,230 and NH16,262) were other substitution occurred in a different clone. Further- used, 5 out of 5 clones yielded the putative Neandertal more, in a subsequent amplification from another ex- sequence (Figure 2, extract C). Thus, while this third tract where both primers (L16,254-H16,379) differed independent extract contains a larger amount of confrom the initial amplification, all four differences located temporary human DNA, probably stemming from laborain the segment included in the first amplifications were tory contamination, it confirms that the putative Neanobserved in all 8 clones sequenced (Figure 5, clones dertal sequence is present in the fossil specimen. B13.1–8). Thus, the retrieval of the putative Neandertal In summary, these experiments indicate that the putasequence is not dependent on the primers used. Fur- tive Neandertal sequence does not originate from a nuthermore, most primer combinations yield a large ex- clear mtDNA insertion and that it is endogenous to the cess of clones representing the putative Neandertal se- fossil. It furthermore falls outside the variation of the quence over clones similar to contemporary human mtDNA gene pool of modern humans (see below). We mtDNA. therefore conclude that it is derived from the mitochonTo further exclude the possibility that the sequence drial genome of the Neandertal individual. may represent a nuclear insertion, primers for the putative Neandertal sequence were constructed that do not amplify human mtDNA. In control experiments where Determination of the Neandertal mtDNA Sequence various amounts of a cloned copy of the putative Nean- The entire sequence of hypervariable region I of the dertal sequence were mixed with human DNA, these mtDNA control region (positions 16,023 to 16,400; Anprimers were able to detect about 20 copies of the derson et al., 1981) was determined. Since the state of cloned sequence in 50 ng of total human DNA, i.e., less preservation of the DNA allowed only short fragments to than one copy per genome equivalent (data not shown). be amplified, this was achieved by several overlapping When these primers were used to amplify DNA isolated amplifications. Furthermore, since the quantitation exfrom 15 Africans, 6 Europeans, and 2 Asians, no amplifi- periments indicated that some amplifications might start cation products were obtained (data not shown),indicat- from single molecules, and thus that misincorporations ing that this sequence is not present in the genome of in early cycles of the amplification might be misintermodern humans. preted as sequence differences (Handt et al., 1996), all To test whether the extraction and amplification of sequence positions were determined from at least two the putative Neandertal sequence is reproducible, an independent amplifications. At five sequence positions, additional independent DNA extraction was performed two amplifications yielded discordant results, i.e., all
eandertal mtDNA Sequence 正正dc正〓啁正正嘿嚯≤ 史牙 g9卫著 西-8 它足定呢明明明定哐定哐 9工8°29Eaa o5o
Neandertal mtDNA Sequence 23 Figures 4. The DNA Sequences of Clones Used to Infer the Sequence of the Hypervariable Region I of the Neandertal Individual Above, the modern human reference sequence (Anderson et al., 1981) is given, below the sequence inferred for the Neandertal individual, numbered according to the reference sequence. The designations and sequences of primers used (reversed and complemented when the letter H occurs in the designations) are given for the first clone of each amplification, except for primers L16,022 (59-CTA AGA TTC TAA TTT AAA CTA TTC CTC T-39) and H16,401 (59-TGA TTT CAC GGA GGA TGG TG-39). For primers L16,209 and H16,271 and further details, see legend to Figure 2. Ambiguities in the sequencing reactions are indicated by standard abbreviations
cerDHTAouofeCenlG Figure 5. The DNA Sequences of Clones Used to Infer the Sequence of the Hypervariable Region I of the Neandertal Individual bove, the modem human reference sequence(Anderson et al., 1981)is given, below the sequence inferred for the Neandertal individual, ter H occurs in the designations) are given for the first clone of each amplification, except for primers L16,022(5'-CTA AGA TTC TAA TTT AAA CTA TTC CTC T-3) and H16, 401(5'-TGA TTT CAC GGA GGA TGG TG-3). For primers L16, 209 and H16, 271 and further details, see legend to Figure 2. Ambiguities in the sequencing reactions are indicated by standard abbreviations clones in one amplification differed at a position from uences occurring in one or more individuals, found all clones in another amplification. For these positions, in 478 Africans, 510 Europeans, 494 Asians, 167 Native clones from at least one more independent amplification Americans and 20 individuals from australia and ocea were sequenced. In all cases, all clones from the sul nia(SMeyer, personal communication). Whereas these quent amplification products carried one of the two modern human sequences differ among themselves by bases at the five positions in question. At 23 positions, an average of 8.0+ 3.1(range 1-24)substitutions, the differences were found between two or more clones difference between the humans and the Neandertal se- amplification or in different amplifica quence is 27. 2+ 2.2(range 22-36)substitutions. Thus, tions In those cases, the sequence present in the major the largest difference observed between any two human ity of clones was scored. Figures 4 and 5 show how 123 sequences was two substitutions larger than the small- clones from 13 amplifications were used to determine est difference between a human and the Neander 379 bp from the Neandertal individual tal. In total, 0.002% of the pairwise comparisons be- tween human mt DNA sequences were larger than the smallest difference between the neandertal and a human the human reference sequence, 27 differences are seen loch Neandertal seque efferent continents, differs by When the Neandertal DNA sequence is compared to utside the heteroplasmic cytosine homopolymer(Be 28.2+ 1.9 substitutions from the European lineages. dall and Sykes, 1995)(Figure 4). Of these 27 differences, 27.1+2.2 substitutions from the African lineages, 27.7+ 24 are transitions, two are transversions, and one repre- 2.1 substitutions from the Asian lineages, 27. 4 1.8 sents an insertion of a single adenosine residue substitutions from the American lineages and 28.3+ The Neandertal sequence was compared to a collec- 3.7 substitutions from the Australian/Oceanic lineages tion of 2051 human and 59 common chimpanzee se- Thus, whereas the Neandertals inhabited the same geo- juences over 360 bp of the sequence determined from graphic region as contemporary Europeans, the ob. the Neandertal (positions 16,024 to 16, 383). Among the served differences between the Neandertal sequence 27 nucleotide differences to the reference sequence and modern Et ns do not indicate that it is more und in this segment, 25 fall among the 225 positions closely related to modern Europeans than to any other hat vary in at least one of the human sequences, and population of contemporary humans. ne of the two remaining positions varies among the When the comparison was extended to 16 common chimpanzees. Thus, the types of differences observed chimpanzee lineages(Figure 6), the number of positions (e. g an excess of transitions over transversions), and mmon among the hum the positions in the Neandertal sequence where they quences was reduced to 333(Morin et al., 1994). This occur, reflect the evolutionary pattern typical of mtDNA reduced the number of human lineages to 986.The aver equences of extant humans and chimpanzees ge number of differences among humans is 8.0+ 3.0 The Neandertal ce was compared to 994 con- (range 1-24), that between humans and the Neandertal temporary human mitochondrial lineages, i.e., distinct 25.6+ 2.2(range 20-34), and that between humans
Cell 24 Figure 5. The DNA Sequences of Clones Used to Infer the Sequence of the Hypervariable Region I of the Neandertal Individual Above, the modern human reference sequence (Anderson et al., 1981) is given, below the sequence inferred for the Neandertal individual, numbered according to the reference sequence. The designations and sequences of primers used (reversed and complemented when the letter H occurs in the designations) are given for the first clone of each amplification, except for primers L16,022 (59-CTA AGA TTC TAA TTT AAA CTA TTC CTC T-39) and H16,401 (59-TGA TTT CAC GGA GGA TGG TG-39). For primers L16,209 and H16,271 and further details, see legend to Figure 2. Ambiguities in the sequencing reactions are indicated by standard abbreviations. clones in one amplification differed at a position from sequences occurring in one or more individuals, found all clones in another amplification. For these positions, in 478 Africans, 510 Europeans, 494 Asians, 167 Native clones from at least one more independent amplification Americans and 20 individuals from Australia and Oceawere sequenced. In all cases, all clones from the subse- nia (S. Meyer, personal communication). Whereas these quent amplification products carried one of the two modern human sequences differ among themselves by bases at the five positions in question. At 23 positions, an average of 8.0 6 3.1 (range 1–24) substitutions, the differences were found between two or more clones, difference between the humans and the Neandertal seeither within one amplification or in different amplifica- quence is 27.2 6 2.2 (range 22–36) substitutions. Thus, tions. In those cases, the sequence present in the major- the largest difference observed between any two human ity of clones was scored. Figures 4 and 5 show how 123 sequences was two substitutions larger than the smallclones from 13 amplifications were used to determine est difference between a human and the Neander- 379 bp from the Neandertal individual. tal. In total, 0.002% of the pairwise comparisons between human mtDNA sequences were larger than the Sequence Comparisons smallest difference between the Neandertal and a human. When the Neandertal DNA sequence is compared to The Neandertal sequence, when compared to the mithe human reference sequence, 27 differences are seen tochondrial lineages from different continents, differs by outside the heteroplasmic cytosine homopolymer (Ben- 28.2 6 1.9 substitutions from the European lineages, dall and Sykes, 1995) (Figure 4). Of these 27 differences, 27.1 6 2.2 substitutions from theAfrican lineages, 27.76 24 are transitions, two are transversions, and one repre- 2.1 substitutions from the Asian lineages, 27.4 6 1.8 sents an insertion of a single adenosine residue. substitutions from the American lineages and 28.3 6 The Neandertal sequence was compared to a collec- 3.7 substitutions from the Australian/Oceanic lineages. tion of 2051 human and 59 common chimpanzee se- Thus, whereas the Neandertals inhabited the same geoquences over 360 bp of the sequence determined from graphic region as contemporary Europeans, the obthe Neandertal (positions 16,024 to 16,383). Among the served differences between the Neandertal sequence 27 nucleotide differences to the reference sequence and modern Europeans do not indicate that it is more found in this segment, 25 fall among the 225 positions closely related to modern Europeans than to any other that vary in at least one of the human sequences, and population of contemporary humans. one of the two remaining positions varies among the When the comparison was extended to 16 common chimpanzees. Thus, the types of differences observed chimpanzee lineages (Figure 6), the number of positions (e.g., an excess of transitions over transversions), and in common among the human and chimpanzee sethe positions in the Neandertal sequence where they quences was reduced to 333 (Morin et al., 1994). This occur, reflect the evolutionary pattern typical of mtDNA reduced the number of human lineages to 986. The aversequences of extant humans and chimpanzees. age number of differences among humans is 8.0 6 3.0 The Neandertal sequence was compared to 994 con- (range 1–24), that between humans and the Neandertal, temporary human mitochondrial lineages, i.e., distinct 25.6 6 2.2 (range 20–34), and that between humans
d chimpanzees, 55.0+ 3.0(range 46-67). Thus, the equences were plotted for each quartet. An example average number of mt DNA sequence differences be- of this analysis is shown in Figure 7b. A different random tween modern humans and the Neandertal is about three subset of 100 human lineages was then chosen and the times that among humans, but about half of that be- analysis repeated; in 40 such analyses, an average of tween modern humans and modern chimpanzee 89%(range 84%93%)of the quartets grouped the two An insertion of a portion of the mitochondrial control uman sequences together. Thus, the phylogenetic anal- region on chromosome 11 has been inferred to repre- yses agree with the pairwise comparisons of sequenc sent an outgroup to the modern human mtDNa gene differences in placing the Neandertal mtDNA sequen pool (Zischler et aL., 1995). When its sequence was com outside the variation of modem human mtDNA pared to the human and Neandertal sequences, 294 positions could be compared and the number of human Age of the Neandertal/Modern Human lineages was reduced to 970. In this case, the differ- mtDNA Ancestor ences between humans and the Neandertal sequence To estimate the time when the most recent ancestral are 25.5+2.1(range 20-34)whereas between humans sequence common to the Neandertal and modern hu and the insertion sequence there are 21.3+ 1.7(range an mtDNA sequences existed, we used an estimated 16-27)differences. Thus, although the distributions of divergence date between humans and chimpanzees of the distances overlap, they suggest that the insertion 4-5 million years ago(Takahata et al., 1995) and cor- of a portion of the mtDNA control region to chromosome rected the observed sequence differences for multiple substitutions at the same nucleotide site(Tamura and dertal and modern human mtDNa gene pools. This is Nei, 1993). This yielded a date of 550, 000 to 690,000 compatible with the notion that the Neandertal years before present for the divergence of the Neander human mtDNA gene pool well before the time of the age of the modern human mtDNA ancestor is estimated most recent common ancestor of human mtdNA using the same procedure, a date of 120,000 to 150,000 years is obtained, in agreement with previous estimates Phylogenetic Analyses (Cann et al. 1987; Vigilant et al., 1991). Although thes To further investigate the relationship of the Neandertal dates rely on the calibration point of the chimpanzee- mtDNA sequence to contemporary human mtDNA varia human divergence and have errors of unknown magni tion, phylogenetic tree reconstructions were performed tude associated with them, they indicate that the age A neighbor-joining tree(Saitou and Nei, 1987)of the 16 of the common ancestor of the Neandertal sequence chimpanzee lineages, the Neandertal sequence, and the and modern human sequences is about four times 986 human lineages was constructed(Figure 7a). This greater than that of the common ancestor of modern e shows the Neandertal sequence diverging prior to human mtDNAs the divergence of the human mtDNA lineages. To esti- mate the support for this relationship, a likelihood map- Rooting of Modern Human mtDNA Gene Pool ping statistic(Strimmer and von Haeseler, 1997)was The phylogenetic tree(Figure 7a)shows the first three used In this analysis, all possible quartets involving the Neandertal sequence, one of the chimpanzee lineage uman branches to be composed of seven African and two representatives out of 100 human lineages(ran- mtDNA sequences, with the first non-African sequences domly selected from the 986 human lineages)were appearing only in the fourth branch. This branching pat analyzed, and the likelihoods for the three possible tern would indicate that the ancestor of the mtDNA gene groupings of the Neandertal, chimpanzee, and human pool of contemporary humans lived in Africa(cf. von Haeseler et aL. 1996). When the statistical support for hese branches was assessed with the likelihood map ping approach using the Neandertal sequence as an ungroup, these three branches were supported by 91% 91%, and 92%of quartets, respectively. When the seven African sequences in the first three branches were tested together with the 684 non-African lineages in the database and the Neandertal sequence, they grouped vith the latter in 97% of all quartets analyzed. Thus, overall, the results suggest an African origin of the hu an mtDNA gene pool, as has been claimed when chim igilant et aL. 1991: Hedges et al. 1992)and a nuclear insertion of the mitochondrial con trol region (Tischler et al. 1995)were used in similar analyses DNA Preservation in the Neandertal Fossil er of sequence differences: Y axis, the percent of Based on its"classical"morphology, the undated Nean pairwise comparison dertal fossil is thought to be between 30,000 and 100,0
Neandertal mtDNA Sequence 25 and chimpanzees, 55.0 6 3.0 (range 46–67). Thus, the sequences were plotted for each quartet. An example average number of mtDNA sequence differences be- of this analysis is shown in Figure 7b. A different random tween modern humans and the Neandertal is aboutthree subset of 100 human lineages was then chosen and the times that among humans, but about half of that be- analysis repeated; in 40 such analyses, an average of tween modern humans and modern chimpanzees. 89% (range 84%–93%) of the quartets grouped the two An insertion of a portion of the mitochondrial control human sequences together. Thus, the phylogenetic analregion on chromosome 11 has been inferred to repre- yses agree with the pairwise comparisons of sequence sent an outgroup to the modern human mtDNA gene differences in placing the Neandertal mtDNA sequence pool (Zischler et al., 1995). When its sequence was com- outside the variation of modern human mtDNA. pared to the human and Neandertal sequences, 294 positions could be compared and the number of human Age of the Neandertal/Modern Human lineages was reduced to 970. In this case, the differ- mtDNA Ancestor ences between humans and the Neandertal sequence To estimate the time when the most recent ancestral are 25.5 6 2.1 (range 20–34) whereas between humans sequence common to the Neandertal and modern hu- and the insertion sequence there are 21.3 6 1.7 (range man mtDNA sequences existed, we used an estimated 16–27) differences. Thus, although the distributions of divergence date between humans and chimpanzees of the distances overlap, they suggest that the insertion 4–5 million years ago (Takahata et al., 1995) and cor- of a portion of the mtDNA control region to chromosome rected the observed sequence differences for multiple 11 may have occurred after the divergence of the Nean- substitutions at the same nucleotide site (Tamura and dertal and modern human mtDNA gene pools. This is Nei, 1993). This yielded a date of 550,000 to 690,000 compatible with the notion that the Neandertal se- years before present for the divergence of the Neander- quence diverged from the lineage leading to the current tal mtDNA and contemporary human mtDNAs. When the human mtDNA gene pool well before the time of the age of the modern human mtDNA ancestor is estimated most recent common ancestor of human mtDNAs. using the same procedure, a date of 120,000 to 150,000 years is obtained, in agreement with previous estimates Phylogenetic Analyses (Cann et al., 1987; Vigilant et al., 1991). Although these To further investigate the relationship of the Neandertal dates rely on the calibration point of the chimpanzee– mtDNA sequence to contemporary human mtDNA varia- human divergence and have errors of unknown magnition, phylogenetic tree reconstructions were performed. tude associated with them, they indicate that the age A neighbor-joining tree (Saitou and Nei, 1987) of the 16 of the common ancestor of the Neandertal sequence chimpanzee lineages, the Neandertal sequence, and the and modern human sequences is about four times 986 human lineages was constructed (Figure 7a). This greater than that of the common ancestor of modern tree shows the Neandertal sequence diverging prior to human mtDNAs. the divergence of the human mtDNA lineages. To estimate the support for this relationship, a likelihood map- Rooting of Modern Human mtDNA Gene Pool ping statistic (Strimmer and von Haeseler, 1997) was The phylogenetic tree (Figure 7a) shows the first three used. In this analysis, all possible quartets involving the human branches to be composed of seven African Neandertal sequence, one of the chimpanzee lineages, mtDNA sequences, with the first non-African sequences and two representatives out of 100 human lineages (ran- appearing only in the fourth branch. This branching pat- domly selected from the 986 human lineages) were tern would indicate that the ancestor of the mtDNA gene analyzed, and the likelihoods for the three possible pool of contemporary humans lived in Africa (cf. von groupings of the Neandertal, chimpanzee, and human Haeseler et al., 1996). When the statistical support for these branches was assessed with the likelihood mapping approach using the Neandertal sequence as an outgroup, these three branches weresupported by 91%, 91%, and 92%of quartets, respectively.When the seven African sequences in the first three branches were tested together with the 684 non-African lineages in the database and the Neandertal sequence, they grouped with the latter in 97% of all quartets analyzed. Thus, overall, the results suggest an African origin of the human mtDNA gene pool, as has been claimed when chimpanzee sequences (Vigilant et al., 1991; Hedges et al., 1992) and a nuclear insertion of the mitochondrial control region (Zischler et al., 1995) were used in similar analyses. Discussion Figure 6. Distributions of Pairwise Sequence Differences among Humans, the Neandertal, and Chimpanzees DNA Preservation in the Neandertal Fossil X axis, the number of sequence differences; Y axis, the percent of Based on its “classical” morphology, the undated Neanpairwise comparisons. dertal fossil is thought to bebetween 30,000 and 100,000
a on-Africans human 1 1 African 892% American 5.4% 4% 4 Africans Neandertal chimp Neandertal igure 7. A Schematic Phylogenetic Tree Relating the Neandertal mtDNA Sequence to 986 Modern Human mtDNA Sequences and Likelihood Mapping Analysis showing the Support for Various Groupings of Neandertal, Human, and Chimpanzee Sequ (a) The tree was rooted with 16 chimpanzee mtDNA lineages. For clarity, only the first five branches without their internal branching structure ut with their geographical states are shown. Numbers on internal branches refer to quartet puzzling probabilities. To calculate these possible combinations of the Neandertal sequence, one of 16 chimpanzee lineages, and two of 100 lineages chosen at random from among (b) For each such quartet of sequences, the likelihoods for each of the three possible phylogenetic arrangements are plotted in a triangle(b upper panel) where the tips indicate absolute support for one of the arrangements (b, lower panel ). The percentage of the quartets favorin the grouping of the Neandertal sequence with the chimpanzee to the exclusion of the two human lineages is found in the upper of the three areas. A total of 40 such analyses with different random sets of human mtDNA lineages were carried out and the average of these is given in the tree. The other internal branches were similarly analyzed. years old(Stringer and Gamble, 1993). It is thus among enhance misincorporations(Hoss et al., 1996a). In addi he oldest specimens for which the chemical stability of tion, the Neandertal extracts contain sequences that are DNA would seem to allow for the retrieval of endogenous probably derived from contemporary humans. This is DNA(Paabo and wilson, 1991; Lindahl, 1993a). Further- not unexpected since the specimen has been exten- more, the extent of amino acid racemization indicates sively handled during 140 years. A further complication hat preservation conditions of the Neandertal fossil is that exogenous as well as endogenous molecules have been compatible with DNA preservation(Poinar et both sometimes carrying substitutions from early cycles aL. 1996). In agreement with this, the quantitation shows of the amplification, may recombine with each other that an extract of 0. 4 g of bone contains about 1000- during the amplification process. Such "jumping PCR 1500 Neandertal mtDNA molecules of length 100 bp. is induced by strand breaks and dNa damage and may Thus, mitochondrial DNA sequences can be retrieved itself introduce sequence changes(Paabo et al. 1990) from the fossil. however this result also indicates that Fortunately, neither misincorporations nor jumping single-copy nuclear DNA sequences, which are a hun- phenomena are expected to show a great specificity for dred -to a thousand-fold less abundant than mtDNA in certain sequence positions. Therefore, even when an ld be impossible to reproducibly amplify from the extracts. This is reminiscent of the situation in occurs in the first cycle of the PCR such that it becomes most other archaeological remains(e.g. Handt et al. represented in all resulting molecules, one will be alerted 1994b,1996) to the problem if two independent amplifications are Several factors complicate etermination o analyzed since they will yield different sequences. A itDNA sequences from the Neandertal fossil. The low third amplification can then be analyzed to determine number of preserved mt DNA molecules poses problems which of the two sequences is reproducible and hence since misincoporations during the initial cycles of the authentic. Taken together, the results( Figures 4 and 5) amplification will become represented in a large fraction show that misincorporations are a fairly frequent phe of the molecules in the final amplification product. Such nomenon in the Neandertal extracts, contributing to vari misincorporations may be particularly likely in the case ation seen at 64 of the 378 sequence positions deter- of ancient DNA, which often contains lesions that can mined. Furthermore, some amplifications start from very
Cell 26 Figure 7. A Schematic Phylogenetic Tree Relating the Neandertal mtDNA Sequence to 986 Modern Human mtDNA Sequences and Likelihood Mapping Analysis Showing the Support for Various Groupings of Neandertal, Human, and Chimpanzee Sequences (a) The tree was rooted with 16 chimpanzee mtDNA lineages. For clarity, only the first five branches without their internal branching structures but with their geographical states are shown. Numbers on internal branches refer to quartet puzzling probabilities. To calculate these, all possible combinations of the Neandertal sequence, one of 16 chimpanzee lineages, and two of 100 lineages chosen at random from among 986 human lineages were analyzed. (b) For each such quartet of sequences, the likelihoods for each of the three possible phylogenetic arrangements are plotted in a triangle (b, upper panel) where the tips indicate absolute support for one of the arrangements (b, lower panel). The percentage of the quartets favoring the grouping of the Neandertal sequence with the chimpanzee to the exclusion of the two human lineages is found in the upper of the three areas. A total of 40 such analyses with different random sets of human mtDNA lineages were carried out and the average of these is given in the tree. The other internal branches were similarly analyzed. years old (Stringer and Gamble, 1993). It is thus among enhance misincorporations (Ho¨ ss et al., 1996a). In addithe oldest specimens for which the chemical stability of tion, the Neandertal extracts contain sequences that are DNA would seem toallow for the retrieval of endogenous probably derived from contemporary humans. This is DNA (Pa¨ a¨bo and Wilson, 1991; Lindahl, 1993a). Further- not unexpected since the specimen has been extenmore, the extent of amino acid racemization indicates sively handled during 140 years. A further complication that preservation conditions of the Neandertal fossil is that exogenous as well as endogenous molecules, have been compatible with DNA preservation (Poinar et both sometimes carrying substitutions from early cycles al., 1996). In agreement with this, the quantitation shows of the amplification, may recombine with each other that an extract of 0.4 g of bone contains about 1000– during the amplification process. Such “jumping PCR” 1500 Neandertal mtDNA molecules of length 100 bp. is induced by strand breaks and DNA damage and may Thus, mitochondrial DNA sequences can be retrieved itself introduce sequence changes (Pa¨ a¨bo et al., 1990). from the fossil. However, this result also indicates that Fortunately, neither misincorporations nor jumping single-copy nuclear DNA sequences, which are a hun- phenomena are expected to show a great specificity for dred- to a thousand-fold less abundant than mtDNA in certain sequence positions. Therefore, even when an most cells, would be impossible to reproducibly amplify amplification starts from a single molecule and an error from the extracts. This is reminiscent of the situation in occurs in the first cycle of the PCR such that it becomes most other archaeological remains (e.g., Handt et al., represented in all resulting molecules, one will bealerted 1994b, 1996). to the problem if two independent amplifications are Several factors complicate the determination of analyzed since they will yield different sequences. A mtDNA sequences from the Neandertal fossil. The low third amplification can then be analyzed to determine number of preserved mtDNA molecules poses problems which of the two sequences is reproducible and hence since misincoporations during the initial cycles of the authentic. Taken together, the results (Figures 4 and 5) amplification will become represented in a large fraction show that misincorporations are a fairly frequent pheof the molecules in the final amplification product. Such nomenon in theNeandertal extracts,contributing tovarimisincorporations may be particularly likely in the case ation seen at 64 of the 378 sequence positions deterof ancient DNA, which often contains lesions that can mined. Furthermore, some amplifications start from very
few or single molecules since amplifications yielded record indicates a likely minimum date completely discordant results at five positions. In con- gence between modern humans and Ne Is of trast, jumping PCR, which is a frequent phenomenon in 50000-300,000 years( Stringer,1997) the ar extracts of some ancient specimens(Handt et al haeological record also puts the divergence between 1994b), does not seem to play a major role in the case modern humans and Neandertals at about 300,000 years of the Neandertal fossil, although some clones( Figure( Foley and Lahr, 1997). a date of over 500,000 years 2, extract B) are likely to be the result of this process for the molecular divergence between Neandertal and human mtDNAs is in excellent agreement with the Implications for Modern Human Origins palaeontological and archaeological record since the Both pairwise sequence comparisons and phylogenetic divergence of genes is expected to predate the diver- gence of populations by an amount that reflects the analyses tend to place the Neandertal mtDNA Sequence level of polymorphism in the ancestral species(Nei outside modern human mtDNA variation. Furthermor the divergence between the Neandertal mtDNA se- 1987). Thus, if the palaeontological and archaeological estimates for the divergence of the Neandertal and hu man populations are accurate, and the mtDNA estimate is estimated to be about four-fold older than the diversity for the molecular divergence is also accurate, this would indicate that the diversity of the mtDNa gene pool in that the diversity among Neandertal mtDNA sequences the ancestral species(presumably Homo erectus)from would have to be at least four times larger than among modern humans in order for other Neandertal sequences which Neandertals and humans evolved. was at least to be ancestral to modern human sequences. Thus, al as great as that of modern h must be emphasized that the above conclusions though based on a single Neandertal sequence, the are based on a single individual sequence; the retrieval present results indicate that Neandertals did not contrib- ute mtDNA to modern humans and analysis of mtDNA sequences from additional These results do not rule out the possibility that Nean Neandertal specimens is obviously desirable. If this proves possible, then the potential exists to address dentals contributed other genes to modern humans. However the view that Neandertals would have contrib several questions concerning Neandertals that hitherto uted little or nothin ng to the modern human ge ne pool could be studied exclusively by morphological and ar- chaeological approaches. For example, the genetic is gaining support from studies of molecula relationship between Neandertal populations in Europ variation at nuclear loci in humans(Hammer, 1995; ind in western Asia could be explored, as could the Armour et al. 1996 Tishkoff et al 1996). It is also in demographic history of Neandertal populations, using agreement with assessments of the degree of morpho- methods that have been applied to investigate the de- logical difference between Neandertal skeletal remains mographic history of modern human populations(Har and modern humans( e.g, Rak, 1993; Zollikofer et al pending et aL., 1993; von Haeseler et al., 1996) 1995; Hublin et al, 1996; Schwartz and Tattersall, 1996) that would classify Neandertals and modern humans as A Cautionary Note separate specie Remains of animals found in association with Neandertal Given the placement of the Neandertal mtDNA se- mains in other parts of Europe have failed to yield quence outside the range of modern human mtDNA amplifiable DNA and/or display levels of amino acid ra- variation, it can be used as an outgroup in phylogenetic cemization that make the prospect of retrieving dNA analyses to assess the geographic origin of the human bleak (A Cooper and H. Poinar, personal communica- mtDNA ancestor. Initial claims that Africa was the most tion). It is therefore possible that the type specimen may likely geographic source of contemporary human mt DNA in containing amplifiable endo genous variation(Cann et al., 1987; Vigilant et al. 1991)were DNA. Thus, we strongly recommend that valuable Nean challenged by subsequent reanalyses that found the dertal specimens should not be subjected to destructive original phylogenetic analyses to be inadequate( Maddi- sampling before the analysis of associated animal fos son,1991; Templeton, 1992; Hedges et al. 1992). How sils, and/or the application of some other method that ever, new methods of phylogenetic analysis have contin requires minimal destruction of specimens (such as ed to support an African origin of human mtDNA amino acid racemization), has yielded evidence that variation( Penny et al. 1995), as has the use of a nuclear dNa may survive in the fossil mtDNA insertion as an outgroup(Tischler et al., 1995) When the Neandertal mt DNA sequence is used to root Experimental Procedures a neighbor joining tree of modern human mtDNA se- quences(Figure 7a), the first three branches consist exclusively of African sequences. The Neandertal mtDNA used were treated with 1 M Hcl equence thus supports a scenario in which modern humans arose recently in Africa as a distinct species ut into a sterile tube for transport to Munich. All subsequent manip and replaced Neandertals with little or no interbreeding ations of the sam Implications for Neandertal Genetics eparate equipment and reagents, UV irradiation, and other It is interesting to compare the mtDNA date for the diver gence between Neandertals and modern humans of 550,000 to 690,000 years ago with dates derived from les of 10 mg of bone powder and varnish were removed by other sources of information. For example, the fossil and were hydrolyzed in 1 ml 6 N HCl at 100C for 24 hr. The
Neandertal mtDNA Sequence 27 few or single molecules since amplifications yielded record indicates a likely minimum date for the divercompletely discordant results at five positions. In con- gence between modern humans and Neandertals of trast, jumping PCR, which is a frequent phenomenon in 250,000–300,000 years (Stringer, 1997), while the arextracts of some ancient specimens (Handt et al., chaeological record also puts the divergence between 1994b), does not seem to play a major role in the case modern humansand Neandertals at about 300,000 years of the Neandertal fossil, although some clones (Figure (Foley and Lahr, 1997). A date of over 500,000 years 2, extract B) are likely to be the result of this process. for the molecular divergence between Neandertal and human mtDNAs is in excellent agreement with the palaeontological and archaeological record since the Implications for Modern Human Origins divergence of genes is expected to predate the diver- Both pairwise sequence comparisons and phylogenetic gence of populations by an amount that reflects the analyses tend to place the Neandertal mtDNA sequence level of polymorphism in the ancestral species (Nei, outside modern human mtDNA variation. Furthermore, 1987). Thus, if the palaeontological and archaeological the divergence between the Neandertal mtDNA se- estimates for the divergence of the Neandertal and hu- quence and the modern human mitochondrial gene pool man populations are accurate, and the mtDNA estimate is estimated to beabout four-fold older than the diversity for the molecular divergence is also accurate, this would of the modern human mtDNA gene pool. This shows indicate that the diversity of the mtDNA gene pool in that the diversity among Neandertal mtDNA sequences the ancestral species (presumably Homo erectus) from would have to be at least four times larger than among which Neandertals and humans evolved, was at least modern humans inorder for otherNeandertal sequences as great as that of modern humans. to be ancestral to modern human sequences. Thus, al- It must be emphasized that the above conclusions though based on a single Neandertal sequence, the are based on a single individual sequence; the retrieval present results indicate that Neandertals did not contrib- and analysis of mtDNA sequences from additional ute mtDNA to modern humans. Neandertal specimens is obviously desirable. If this These results do not rule out the possibility that Nean- proves possible, then the potential exists to address dertals contributed other genes to modern humans. several questions concerning Neandertals that hitherto However, the view that Neandertals would have contrib- could be studied exclusively by morphological and ar- uted little or nothing to the modern human gene pool chaeological approaches. For example, the genetic is gaining support from studies of molecular genetic relationship between Neandertal populations in Europe variation at nuclear loci in humans (Hammer, 1995; and in western Asia could be explored, as could the Armour et al., 1996; Tishkoff et al., 1996). It is also in demographic history of Neandertal populations, using agreement with assessments of the degree of morpho- methods that have been applied to investigate the de- logical difference between Neandertal skeletal remains mographic history of modern human populations (Har- and modern humans (e.g., Rak, 1993; Zollikofer et al., pending et al., 1993; von Haeseler et al., 1996). 1995; Hublin et al., 1996; Schwartz and Tattersall, 1996) that would classify Neandertals and modern humans as A Cautionary Note separate species. Remains of animalsfound in association with Neandertal Given the placement of the Neandertal mtDNA se- remains in other parts of Europe have failed to yield quence outside the range of modern human mtDNA amplifiable DNA and/or display levels of amino acid ravariation, it can be used as an outgroup in phylogenetic cemization that make the prospect of retrieving DNA analyses to assess the geographic origin of the human bleak (A. Cooper and H. Poinar, personal communicamtDNA ancestor. Initial claims that Africa was the most tion). It is therefore possible that the type specimen may likely geographic source of contemporary human mtDNA be fairly unique in containing amplifiable endogenous variation (Cann et al., 1987; Vigilant et al., 1991) were DNA. Thus, we strongly recommend that valuable Neanchallenged by subsequent reanalyses that found the dertal specimens should not be subjected to destructive original phylogenetic analyses to be inadequate (Maddi- sampling before the analysis of associated animal fosson, 1991; Templeton, 1992; Hedges et al., 1992). How- sils, and/or the application of some other method that ever, new methods of phylogenetic analysis have contin- requires minimal destruction of specimens (such as ued to support an African origin of human mtDNA amino acid racemization), has yielded evidence that variation (Penny et al., 1995), as has the use of a nuclear DNA may survive in the fossil. mtDNA insertion as an outgroup (Zischler et al., 1995). When the Neandertal mtDNA sequence is used to root Experimental Procedures a neighbor joining tree of modern human mtDNA se- Sampling quences (Figure 7a), the first three branches consist Protective clothing was worn throughout the sampling procedure. exclusively of African sequences. The Neandertal mtDNA Instruments used were treated with 1 M HCl followed by extensive sequence thus supports a scenario in which modern rinsing in distilled water. After removal, the sample was immediately humans arose recently in Africa as a distinct species put into a sterile tube for transport to Munich. All subsequent manipulations of the sample, and experimental procedures prior to cycling and replaced Neandertals with little or no interbreeding. of PCRreactions, were carried out in laboratories solelydedicated to the analysis of archaeological specimens, whereprotective clothing, separate equipment and reagents, UV irradiation, and other mea- Implications for Neandertal Genetics sures to minimize contamination are used routinely. It is interesting to compare the mtDNA date for the divergence between Neandertals and modern humans of Amino Acid Analysis 550,000 to 690,000 years ago with dates derived from Samples of 10 mg of bone powder and varnish were removed by other sources of information. For example, the fossil drilling and were hydrolyzed in 1 ml 6 N HCl at 1008C for 24 hr. The
re processed and analysed by se phase HPLC mtDNA se e from base 16. 190 to 16. 290 but lacks 12 bases (16,210to16,221) determined by comparison to standards after subtrar ure by alkaline lysis(Ausubelet al., 1995)and the DNA concentration acid t determined in a parallel with t DNA Extraction 5'-GTA GAT TTG TTG ATA TCC TAG TGG GTG TAA-3)specific for Bone pieces were removed using a drill saw. Subsequently, the more treated by soaking in a 10% solution of Naclo for 10 s, and Sequenc ouble-distilled and UV-irradiated H O. The samples were In total, 27 amplifications of the mtDNA control region were per hen ground to powder in a Spex Mill(Edison, N)filled with liquid formed from the Neandertal spe itrogen. the powder was incubated in 1 ml of 0.5 MEDTA (pH 8.0). tions yielded exclusively Neandertal sequences, whereas 6 ampli 5% sarkosyl on a rotary wheel at ambient temperature for 40 hi reference se well as Neandertal Figures 2, 4. tinued at 3]C for another 40 hr. Tissue remains were and 5 give all Neandertal sequences determined except those from nol, phenol/chloroform, and chloroform/isoamyl alcohol as de- yielded exclusively sequences similar to the human reference se- scribed (Ausubel et aL., 1995). The aquaeous phase was concen- quence(not shown). In these cases, the primers used turned out to trated by centrifugal dialysis using Microcon-30 microconcentrators lave one to four mismatches to the Neandertal sequence and thus (Amicon, Beverly, MA)and incubated with 40 ul silica suspension to select against the latter. The Neandertal sequence was inferred in 1 ml of 5 M guanidinium isothiocyanate, 0.1 M Tris-HCI (pH 7. 4 using the sequences shown in Figures 2, 4, and 5. for 15 min on a rotary wheel at ambient temperature as de The Neandertal sequence was compared to an mtDNA sequence ss and Paabo, 1993). The silica was collected by centrifugation database(S Meyer, personal communication). Except for the scor ng of variable deletions, or ambiguities occur were used. In the case of the 16 at 56 C. The eluates were pooled, aliquoted, and stored at chimpanzee lineages, four are reported (Morin et al., 1994) with 20C. An extraction control to which no bone powder was added in ambiguity at position 16, 049, and one each with ambiguities at was processed in parallel with each sample extraction. determined using unpublished software by A von Haeseler. Max mum likelihood distances and phylogenetic trees A lower reaction mixture of 10 ul (67 mM Tris-HCI [pH 8.8]. 2 mM using the PHYLiP package, version 3. 5(Felsenstein, 1994),assum- MgCh,, 2 mg/ml BSA, 2 uM of each primer and 0. 25 mM of each ing a transition/transversion ratio of 20. Support for internal n step by a w mixture(67 mM Tris-HCI (pH 8.8]. 2 mM Zle 3.0(Strimmer and von Haeseler, 1997). Briefly, this algorith MgCI, 0.375 units Taq DNA polymerase [ Perkin Elmer, Cetus], 5 ul analyzes all possible quartets yo from one side bone extract). Forty cycles of PCR(15 s at 92C, 1 min at 55 of an internal branch, two from the other. In each case, enough 60C, 1 min at 72 C)were carried out in an M research PTC 20 subsamples of sequences were analyzed to ensure wit cycler. Ten microliters of the reactions was electrophoresed in 3% bility that any particular sequence was sampled at least once agarose gels, stained with ethidium bromide, and visualized by Uv For dating, the Tamura-Nei algorithm(Tamura and Nei, 1993)as plemented in Puzzle 3.0 was used to estimate 9 classes of substi- he lanes where control amplifications were electrophoresed, w ution rates, rate heterogeneity parameters, transition/transversion cut out of the gels, melted in 100 ul double-distilled H,0 and shock ratios, pyrimidine/purine transition ratios and nucleotide frequen frozen in liquid nitrogen. After thawing during centrifugation, 5 ul of cies for 5 datasets consisting of 100 random human sequences The average values for these parameters were then used to calculate 8.8]. 2 mM MgCl2, 1 mg/ μ M of each prim the distances within and between species. 0. 125 mM of each dNTP, 0.75 units Taq DNA polymerase). Thirty ycles identical to the initial amplification, except for an increase Acknowledgments of 3.C in annealing temperature, were performed. If primer dimers or nonspecific bands were visible upon gel electrophoresis, reampl We are indebted to Drs F. G. Zehnder and H -E. Joachim(Rhei- fication products were gel purified prior to cloning. Alternatively, 10 nisches Landesmuseum Bonn)for permission to remove the sample ul of the reamplification vo to H Ludtke and M. Schultz for support and advice in the samplin merase(New England Biolabs ing to the supplier's pro- process: to s. Meyer and K Strimmer for help with con Biotech. u es:to w. Schartau for oligonucleotide synthesis: to M. Beutels- sala, Sweden) vector in the presence of 10 units of Smal at ambient acher, H. Frohlich, A. Greenwood, R. F. Grill, M. HOSS, T. Merritt, temperature for 16 hr E. coli SURE (Stratagene, La Jolla, CA)were H. Poinar, L. Vigilant, and H. Tischler for discussions and help: to the Deutsche Forschungsgemeinschaft(Pa 452/3-1). the Boehring in 1 ml SoC medium(Ausubel et al. 1995)for 20-25 min before gelheim Fonds(M. K ), and the N ating on selective IPTG/X-gal S )for financial support. R. W. S especially thanks his late Ph.D. White colonies were transferred into 12.5 ul PCR reactions (con- supervisor W. Taut tents as in reamplifications)with"M13 unive Received April 24, 1997; revised June 16, 1997 72C) were carried out ar inserts of the expected size were identified by agarose gel electrophoresis References Of these PCR products, 1.5 ul was sequenced with the Thermo ncing reactions were loaded onto a 6.5% R, Drouin, J, Eperon, I.C., Nierlich, D P, Roe, B.A. Sanger, F. denaturing polyacrylamide gel and analysed on an A L.F. automated etal. (1981). Sequence and organization of the human mitochondrial sequencer(Pharmacia Biotech, Uppsala, Sweden) genome. Nature 290, 457-474 Armour, J.A. L, Anttinen, T- May, C A, Vega, E.E., Sajantila, A, Kidd Kidd, K.K., Bertranpetit atellite diversity sup ecent African origin for modern 1996)and cloned into pUC18. This molecule matches the Neandertal humans. Nature genet. 13, 154-1
Cell 28 hydrolysates were processed and analysed by reverse phase HPLC mtDNA sequence from base 16,190 to 16,290 but lacks 12 bases as described (Poinar et al., 1996). Amounts of the amino acid enanti- (16,210 to 16,221). The plasmid was purified from an overnight culomers were determined by comparison to standards after subtrac- tureby alkaline lysis (Ausubel et al., 1995) and the DNA concentration tion of the amino acid content determined in a negative control determined by UV absorbance at 260/280 nm. The standard and processed in parallel with the samples. Neandertal DNA extracts were used in amplifications with two primers (NL16,209: 59-CCC CAT GCT TAC AAG CAA GC-39, NH16,262: DNA Extraction 59-GTA GAT TTG TTG ATA TCC TAG TGG GTG TAA-39) specific for Bone pieces were removed using a drill saw. Subsequently, the the Neandertal sequence. surface of the pieces was removed with a grinding bit and furthermore treated by soaking in a 10% solution of NaClO for 10 s, and Sequence Analyses rinsing in double-distilled and UV-irradiated H In total, 27 amplifications of the mtDNA control region were per- 2O. The samples were then ground to powder in a Spex Mill (Edison, NJ) filled with liquid formed from the Neandertal specimen. Of these, twelve amplificanitrogen. The powder was incubated in 1 ml of 0.5 M EDTA (pH 8.0), tions yielded exclusively Neandertal sequences, whereas 6 amplifi- 5% sarkosyl on a rotary wheel at ambient temperature for z40 hr. cations contained sequences similar to the contemporary human Ten microliters of proteinase K (10 mg/ml) were then added and the reference sequence as well as Neandertal sequences. Figures 2, 4, incubation continued at 378C for another 40 hr. Tissue remains were and 5 give all Neandertal sequences determined except those from removed by centrifugation and the supernatant extracted with phe- the quantitation reaction (Figure 3). In addition, 9 amplifications nol, phenol/chloroform, and chloroform/isoamyl alcohol as de- yielded exclusively sequences similar to the human reference sescribed (Ausubel et al., 1995). The aquaeous phase was concen- quence (not shown). In these cases, the primers used turned out to trated by centrifugal dialysis using Microcon-30 microconcentrators have one to four mismatches to the Neandertal sequence and thus (Amicon, Beverly, MA) and incubated with 40 ml silica suspension to select against the latter. The Neandertal sequence was inferred in 1 ml of 5 M guanidinium isothiocyanate, 0.1 M Tris-HCl (pH 7.4) using the sequences shown in Figures 2, 4, and 5. for 15 min on a rotary wheel at ambient temperature as described The Neandertal sequence was compared to an mtDNA sequence (Ho¨ ss and Pa¨ a¨ bo, 1993). The silica was collected by centrifugation database (S. Meyer, personal communication). Except for the scorand washed twicewith 1 ml 70% ethanol and once with 1 mlacetone, ing of variable positions, only human sequences where noinsertions, prior to air-drying. DNA was eluted into two aliquots of 65 ml TE (pH deletions, or ambiguities occur were used. In the case of the 16 8.0) at 568C. The eluates were pooled, aliquoted, and stored at chimpanzee lineages, four are reported (Morin et al., 1994) with 2208C. An extraction control to which no bone powder was added an ambiguity at position 16,049, and one each with ambiguities at was processed in parallel with each sample extraction. positions 16,063 and 16,064. Pairwise sequence differences were determined using unpublished software by A. von Haeseler. MaxiPCR, Cloning, and Sequencing mum likelihood distances and phylogenetic trees were computed A lower reaction mixture of 10 m using the PHYLIP package, version 3.5 (Felsenstein, 1994), assum- l (67 mM Tris-HCl [pH 8.8], 2 mM MgCl ing a transition/transversion ratio of 20. Support for internal 2, 2 mg/ml BSA, 2 mM of each primer and 0.25 mM of each dNTP) was separated prior to the first denaturation step by a wax branches was tested by likelihood mapping using the program Puzlayer from a 10 ml upper mixture (67 mM Tris-HCl [pH 8.8], 2 mM zle 3.0 (Strimmer and von Haeseler, 1997). Briefly, this algorithm MgCl analyzes all possible quartets of sequences, two from one side 2, 0.375 units Taq DNA polymerase [Perkin Elmer, Cetus], 5 ml bone extract). Forty cycles of PCR (15 s at 928C, 1 min at 558C or of an internal branch, two from the other. In each case, enough 60 subsamples of sequences were analyzed to ensure with 95% proba- 8C, 1 min at 728C) were carried out in an MJ research PTC 200 cycler. Ten microliters of the reactions was electrophoresed in 3% bility that any particular sequence was sampled at least once. agarose gels, stained with ethidium bromide, and visualized by UV For dating, the Tamura-Nei algorithm (Tamura and Nei, 1993) as implemented in Puzzle 3.0 was used to estimate 9 classes of substi- transillumination. The bands, as well as the corresponding areas of tution rates, rate heterogeneity parameters, transition/transversion the lanes where control amplifications were electrophoresed, were ratios, pyrimidine/purine transition ratios and nucleotide frequen- cut out of the gels, melted in 100 ml double-distilled H2O and shock cies for 5 datasets consisting of 100 random human sequences. frozen in liquid nitrogen. After thawing during centrifugation, 5 ml of The average values for these parameters werethen used to calculate the supernatant was used in 50 ml amplification reactions (67 mM Tris-HCl [pH 8.8], 2 mM MgCl the distances within and between species. 2, 1 mg/ml BSA, 1 mM of each primer, 0.125 mM of each dNTP, 0.75 units Taq DNA polymerase). Thirty Acknowledgments cycles identical to the initial amplification, except for an increase of 38C in annealing temperature, were performed. If primer dimers or nonspecific bands were visible upon gel electrophoresis, reampli- We are indebted to Drs. F. G. Zehnder and H.-E. Joachim (Rheinisches Landesmuseum Bonn) for permission to remove the sample; fication products were gel purified prior to cloning. Alternatively, 10 ml of the reamplification volume were directly treated with T4 DNA to H. Lu¨ dtke and M. Schultz for support and advice in the sampling polymerase (New England Biolabs) according to the supplier’s pro- process; to S. Meyer and K. Strimmer for help with computer analyses; to W. Schartau for oligonucleotide synthesis; to M. Beutels- tocol and ligated into a SmaI-cut pUC18 (Pharmacia Biotech, Uppsala, Sweden) vector in the presence of 10 units of SmaI at ambient pacher, H. Fro¨ hlich, A. Greenwood, R. F. Grill, M. Ho¨ ss, T. Merritt, temperature for 16 hr. E. coli SURE (Stratagene, La Jolla, CA) were H. Poinar, L. Vigilant, and H. Zischler for discussions and help; to the DeutscheForschungsgemeinschaft (Pa 452/3–1), the Boehringer transformed by electroporation using half of the ligation and grown Ingelheim Fonds (M. K.), and the National Science Foundation (M. in 1 ml SOC medium (Ausubel et al., 1995) for 20–25 min before plating on selective IPTG/X-gal agar plates. S.) for financial support. R. W. S. especially thanks his late Ph.D. White colonies were transferred into 12.5 m supervisor W. Taute (University of Cologne). l PCR reactions (contents as in reamplifications) with “M13 universal” and “M13 reverse” primers. After 5 min at 92 Received April 24, 1997; revised June 16, 1997. 8C, 30 cycles of PCR (30 s at 908C, 1 min at 508C, 1 min at 728C) were carried out and clones with inserts of the expected size were identified by agarose gel electrophoresis. References Of these PCR products, 1.5 ml was sequenced with the Thermo Sequenase kit (Amersham, UK) according to the supplier’s instruc- Anderson, S., Bankier, A.T., Barrell, B.G., de Bruijn, M.H.L., Coulson, tions, and half of the sequencing reactions were loaded onto a 6.5% A.R., Drouin, J., Eperon, I.C., Nierlich, D.P., Roe, B.A., Sanger, F., denaturing polyacrylamide gel and analysed on an A.L.F. automated et al. (1981). Sequence and organization of the human mitochondrial sequencer (Pharmacia Biotech, Uppsala, Sweden). genome. Nature 290, 457–474. Armour, J.A.L., Anttinen, T., May, C.A., Vega, E.E.,Sajantila, A., Kidd, Quantitative PCR J.R., Kidd, K.K., Bertranpetit, J., Pa¨ a¨ bo, S., and Jeffreys, A.J. (1996). A competitor standard was constructed (Fo¨rster 1994; Handt et al. Minisatellite diversity supports a recent African origin for modern 1996) and cloned into pUC18. This molecule matches the Neandertal humans. Nature Genet. 13, 154–160