Dating of the Human-Ape Splitting by a Molecular Clock of Mitochondrial DNA

J Mol Evol (1985) 22:160-174 Journal of Molecular Evolution ~) Springer-Verlag 1985

Masami Hasegawa, ~ Hirohisa Kishino, 1 and Taka-aki Yano 2

1 The Institute of Statistical Mathematics, 4-6-7 Minami-Azabu, Minato-ku, Tokyo 106, Japan 2 College of Arts and Sciences, Showa University, Fuji-Yoshida, Yamanashi 403, Japan

Summary. A new statistical method for estimating Introduction divergence dates of species from DNA sequence data by a molecular clock approach is developed. This When humans and apes separated during evolution method takes into account effectively the informa- is still a matter of controversy. The fossil record, tion contained in a set of DNA sequence data. The of course, can provide relevant data, but it does not molecular clock of mitochondrial DNA (mtDNA) provide conclusive evidence, because the data caO was calibrated by setting the date of divergence be- be interpreted in several ways. The molecular record tween primates and ungulates at the Cretaceous- can provide additional powerful material to solve Tertiary boundary (65 million years ago), when the this problem. extinction of dinosaurs occurred. A generalized least- Because of the approximate constancy of the rate squares method was applied in fitting a model to of change in informational macromolecules, it has mtDNA sequence data, and the clock gave dates of been suggested that they can serve as an evolution" 92.3 ___ 11.7, 13.3 + 1.5, 10.9 +_ 1.2,3.7 +__ 0.6, and ary clock allowing us to date the divergence times 2.7 _ 0.6 million years ago (where the second of of extant organisms (Zuckerkandl and Pauling 1962, each pair of numbers is the standard deviation) for 1965; Dickerson 1971; Wilson et al. 1977). This the separation of mouse, gibbon, orangutan, gorilla, constancy is consistent with the neutral theory of and chimpanzee, respectively, from the line leading molecular evolution (Kimura 1968, 1983; Kimura to humans. Although there is some uncertainty in and Ohta 1974). Since the pioneering work of Saric.la the clock, this dating may pose a problem for the and Wilson (1967), many researchers have estl" widely believed hypothesis that the bipedal creature mated the divergence time between humans and the Australopithecus afarensis, which lived some 3.7 African apes using molecular clock approaches (SaP million years ago at Laetoli in Tanzania and at Had- ich and Cronin 1976, 1977; Andrews and Cronila ar in Ethiopia, was ancestral to man and evolved 1982; Sibley and Ahlquist 1984). In spite of the after the human-ape splitting. Another likelier pos- diverse materials and methods used, their results sibility is that mtDNA was transferred through uniformly show an apparently recent divergence of hybridization between a proto-human and a proto- less than 8 million years (Myr) ago between humans chimpanzee after the former had developed bipedal- and the African apes, which indicates that Rarna" ism. pithecus, which lived some 8-14 Myr ago, cannot have been an ancestor of humans that evolved after Key words: Evolution of hominoids -- Phyloge- the human-ape separation (Andrews 1982; PilbeaOa netic position of Australopithecus afarensis -- In- 1982, 1984; Ciochon and Corruccini 1983 ). It is noVr terspecies transfer of mitochondrial DNA apparent that the molecular record can tell us mucl~ about the dates of branching during hominoid evO" lution. The previous molecular clock studies were based Offprint requests to: M. Hasegawa on immunological distances (Sarich and WilsOn 161 1967; Sarich and Cronin 1976, 1977), DNA hy- rilla, orangutan, and gibbon (Brown et al. 1982). bridization (Sarich and Cronin 1977; Sibley and This segment contains the genes for three tRNAs Ahlquist 1984), restriction endonuclease mapping and parts of two proteins. The data set used in the ~ DNA (Brown et al. 1979), protein present study is composed of the 896-nucleotide electrophoresis (Sarich and Cronin 1976; Nozawa sequences from the above-mentioned five species of et al. 1982), and amino acid sequencing (Goodman Hominoidea and the sequences of the corresponding et al. 1983). Although these methods were powerful regions from bovine and mouse (L-strand of enough to exclude the possibility that Rarnapithecus mtDNA). These data provide us with the oppor- Was ancestral to humans and evolved after the hu- tunity to date the divergence events during the evo- man-ape splitting, they provided only a rough es- lution of the Hominoidea by a more reliable method timate of the date of the separation. than has been used before. These methods estimated genetic distances in- The rate of synonymous substitutions in DNA directly, and not on the basis of statistical models. coding for proteins is much higher than both that They therefore contained some uncertainty, and the of amino acid-altering substitutions (Kimura 1977; amount of error inherent in the estimates could not Brown et al. 1982; Miyata et al., 1982) and that of be evaluated in a proper way. Furthermore, since substitutions in tRNA genes. The rates of amino raost of these previous methods did not take account acid-altering substitutions and of tRNA substitu- of the effect of multiple changes in a site, their es- tions have been approximately the same during the timates of the divergence date are biased in favor evolution of animal mtDNA (Brown et al. 1982). ~176 more ancient than the actual one when a more This is in sharp contrast with the situation for nu- .distant splitting is taken as a reference. Therefore, clear DNA, in which tRNA genes are much more tn a preliminary report, we developed a statistical conservative than are most of the genes for proteins rnethod that gives genetic distances by direct com- (Hasegawa et al. 1984b). Since synonymous substi- Parison between mitochondrial DNA (mtDNA) se- tutions are confined mostly to the third codon po- quences, and obtained a more reliable estimate of sitions of protein genes, we divide the nucleotide luhe timing of the divergence events during the evo- sites into two classes: Class 1 sites are third codon tion of the Hominoidea (Hasegawa et al. 1984a). positions, and class 2 sites are first and second codon Our estimate was heavily dependent on the as- positions and sites in tRNA genes. These two classes ~UraPtion that the divergence between bovines and of sites are treated separately in the statistical model ~rirrtates occurred 90 Myr ago (Dickerson 1971; Sar- presented in this paper. 19h and Cronin 1976; Simons I976; Wilson et al. 77; Goodman et al. 1983). However, we now know t no convincing fossils of the living orders of Phylogenetic Relationships Among the lalacental mammals have been found from the Cre- Hominoidea taceous period (Novacek 1982; Savage and Russell 1983). Also, the presumed holocaust that occurred In analyzing the data, it must be taken into account at the end of the Cretaceous (Alvarez et al. 1980, that transition (A ~ G, T ~ C) has greatly predom- 1984; Raup and Sepkoski 1984), some 65 Myr ago, inated over transversion (A,G ~ T,C) in the evo- raay have been responsible for starting a new ra- lution of animal mtDNA (Brown and Simpson 1982; diation of placental mammals (Allan C. Wilson, per- Brown et al. 1982). We therefore counted the num- SOnal communication). Therefore, it seems likely bers of transition- and transversion-type differences that the divergence between bovines and primates between species in class I and class 2 sites sepa- ~CUrred as recently as 65 Myr ago. In this paper, rately, as shown in Table 1 (Hasegawa et al. 1984a). we Present full details of our method, and give es- It is remarkable that the number of transition-type tiraates of divergence times among the Hominoidea differences in class 1 sites between human and chim- ~ using a recalibrated molecular clock based panzee is nearly the same as that between human ot~ the revised reference time. and mouse. This means that a considerable number of multiple transitions have accumulated at these sites, even when we compare any pair of closest Mitochondrial DNA Sequence Data relatives in the present data set. Transition at the third codon position (class 1 site) is always synon- The rntDNAs of human (Anderson et al. 1981), ymous in the genetic code of mammalian mito- b~vine (Anderson et al. t982), and mouse (Bibb et chondria (BarreU et al. 1979; Anderson et al. 1981, al. 1981), each of which is about 16,500 nucleotides 1982; Bibb et al. 1981). ill length, have been completely sequenced. Brown The transversion-type differences in Table 1 and a~d his coworkers sequenced a stretch of 896 nu- other evidence indicate that of the living hominoids, Cleotides in mtDNAs from human, chimpanzee, go- gibbons separated first and orangutans second from 162

Table l. Numbers of transition- (upper right hal0 and transversion- (lower left half) type nucleotide differences among mammaliaa mtDNAs

1 2 3 4 5 6 i Mouse Bovine Gibbon Orang. Gorilla Chimp. Human SOVr~ 1 Mouse 68 (39) 81 (53) 81 (48) 87 (46) 79 (50) 79 (51) 0.119 (0.206) 2 Bovine 91 (82) 80 (42) 81 (44) 93 (52) 85 (61) 86 (57) 0.128 (0.221) 3 Gibbon 83 (83) 69 (71) 57 (59) 65 (59) 61 (64) 59 (58) 0.09t (0.259) 4 Orang. 90 (85) 65 (65) 18 (34) 64 (52) 59 (60) 55 (53) 0.089 (0.237) 5 Gorilla 85 (77) 72 (67) 19 (26) 15 (18) 28 (58) 32 (52) 0.045 (0.237) 6 Chimp. 86 (79) 71 (67) 18 (26) 16 (18) 5 (4) 24 (50) 0.036 (0.216) Human 89 (77) 70 (67) 19 (26) 15 (20) 4 (4) 3 (2) V(O/r~ O. 131 O. 104 0.028 0.023 0.007 0.005 (0.347) (0.291) (0.121 ) (0.080) (0.017) (0.009) Number in parentheses is for the third codon positions of protein-coding regions (class 1 sites; 232 nucleotides) and the number preceding it is for the rest of the sites (class 2 sites; 667 nucleotides). From Hasegawa et al. (1984a). See text for explanation ofS(~ rl and Vt0/ri the line leading to humans (Goodman 1962, 1963; DNA hybridization (Sibley and Ahlquist 1984), bY Zihlman et al. 1978; Fends et al. 1981a; Andrews hemoglobin sequences (Goodman et al. 1983), and and Cronin 1982; Brown et al. 1982; Sibley and by extensive comparison of high-resolution banding Ahlquist 1984), and that primates are related more patterns of the chromosomes (Yunis and PrakaSta closely to bovines than to the mouse (McKenna 1982). We tentatively adopt this tree topology ia 19 7 5; Eisenberg 19 81). However, the branching or- estimating divergence times in the Hominoidea. der among human, chimpanzee, and gorilla is controversial. Templeton (1983) has developed an algorithm for a nonparametric test for comparing alternative phylogenies obtained from restriction A Statistical Model endonuclease cleavage site data, and has applied it to the mtDNA data from hominoids. His conclusion Let us consider s homologous nucleotide sequenCeS was that the chimpanzee and gorilla separated after that consist of r nucleotide sites of a homogeneous the divergence of humans. Because his analysis in- class (either class 1 or class 2). For the data set volved many synonymous transitions, a consider- analyzed in this work, s = 7, r~ = 232 (class 1 sites), able number of which represent multiple transitions, and r2 = 667 (class 2 sites); sites that experienced his conclusion may not be correct. In fact, nine of deletion or insertion are included, but deletion-ila" the variations in the data used by him are in the sertion events are not taken into account in our protein-coding region in our data set. Seven of them analysis. A basic assumption is that each site changes involve transitions at third codon positions, one of homogeneously and independently of others; that the remaining two involves transitions at a first po- is, the probability of nucleotide substitution has a~ sition, and the other involves transitions at a second independently identical distribution (i.i.d.). A rarl- position. dom variable is represented by (x~ .... , xs), in which To clarify the phylogenetic relationships among each component is T, C, A, or G, and the number the Hominoidea, we applied the maximum likeli- of possible states is 4 S. Our purpose is to parametrize hood method developed by Felsenstein (1981) to P(xl = il ..... xs = is) ---- qil ... is our data set (Hasegawa and Yano 1984). The meth- (i~ ..... is = T, C, A, G) od originally assumed that transitions and transversions occur at the same rate. This assumption is based on a statistical model and to estimate diver" invalid in animal mtDNA. Therefore, we separated gence times among the extant hominoids. transversion from transition, and examined only the We denote by nit i, the number of sites that former in calculating the maximum likelihood es- have a value of(il ..... is). This follows the multi" timate. The topology of the maximum likelihood nomial (4~-nomial) distribution tree (Fig. 1) shows the chimpanzee as the unique Pol(n; qi~.., i,, it,..., is = T, C, A, G) closest relative of humans among extant apes. Al- though the branching order among humans and the and represents the most detailed information aboOt African apes is confident only at 4.4% risk level by the data under the basic assumption of an i.i.d. Tla~ this analysis, the human--chimpanzee grouping has average and the covariance of these statistics afe been suggested also by a single-copy nuclear DNA- given by the following formulae: 163

7. Man where rrj is the stationary composition of base j. In our data set, ~-a-it) = 0.169, ~rco) = 0.429, a'Ao) = 6. Chimpanzee 0.364, and 7to~ = 0.038 for class 1 sites; and 7rx~2) = 5. Gorilla 0.297, 7rca) = 0.267, ~rA(2) = 0.310, and 7rG~2) = 0.126 for class 2 sites. This model is justified because the base composition of animal mtDNA is highly biased 4. Orangutan (particularly, G is scarce in the L-strand), and because the asymmetry of the substitution frequencies t 2 = 65Myr 3. Gibbon is in accord with the bias in base composition (the A --. G frequency is much lower than the G --, A) (Aquadro and Greenberg 1983). Our model is a gen- eralization of the models of Kimura (1980) and of Felsenstein (1981). Kimura's model corresponds to the case of lrx = 7rc = ~'A = ~rG = Va in Eq. (2), and Felsenstein's model corresponds to the case of a = #. Since transition predominates over transversion, Felsenstein's model is apparently inadequate for animal mtDNA. Furthermore, because the base composition of animal mtDNA is highly biased, Ki- mura's model does not fit the data. This will be further shown for the class 1 sites later in this paper. 2. Bovine The substitution probability matrix for an infinitesimally short interval of time can be written as

I. Mouse T C A G big. 1. Phylogeny inferred from the mtDNA sequences by a C aTradt 1 -- (Ct~ T + ~WA /~rAdt /~rodt maximum likelihood method developed by Felsenstein (1981). + B*ro)dt In calibrating our molecular clock, the date of divergence between P(dt) = r /5~rcdt 1 - (ttrr + fl:rr t~rodt the primates and bovines (node 2) was taken to be 65 Myr ago + flrc)dt #Trrdt flrcdt ar^dt 1 -- (a~rA + fir r + fl*rc)dt --- I + Adt (3) E{nil ...i,} = rqi,---is (~Ov{nil ...i,, nit,...is,} = r(6i, ...i,; if...is'qit ...is (la) For an arbitrary time interval t, the function P(t) -- qit.., isqil'.., is') (lb) satisfies the Chapman-Kolmogorov equation Where 6it ... i,; it,.., i,, equals 1 when it = it', .... i, = P(t + dt) = P(t)P(dt) 1~', and 0 otherwise. = P(0(I + AdO We cannot handle Eqs. (la) and (lb) as they are, This equation is a mathematical manifestation of because the number of states increases explosively the Markovian nature of the process. Therefore, we as s increases. Therefore, we reduce the data to dif- get ferences, and compare the differences with a probability distribution to which they conform. dP(t) _ P(t)A dt ,4 Stationary Markov Model Since P(0) = I, we have The probability that a given site is variable is de- P(t) = e tA (4) noted by f, which means that the probability of its To carry out our analysis, it is necessary to ex- being nonvariable is 1 - f. Each variable site evolves plicitly determine the individual substitution prob- aCCording to a Markov process in which a base i (T, ability Pij(0 by using the specific value decompo- C, A, or G) is replaced by another base j in an sition of the right-hand side of Eq. (4). By infinitesimally short interval of time, dt, with a decomposing A as larobability of Pij(t), as follows: 4 Pij(dt) = Pr(x(t + dt) = j Ix(t) = i) A = ~ XiUiXi' i-I = Ia~'jdt (for transition) (2) [#rjdt (for transversion) we have 164

4 Pr(x(t) = i, y(t) = j) etA = ~ exp(t)kl)uivi' i=l = Pr(x(t) = i, y(t) = j [variable site) Pr(variable site) where det(hiI - A) = 0, Aui = ~i~i, A'~i = Xivi, (7) = f ~ 7rtPti(t)Pej(t) (u~, vj) = ~0 for i, j = 1, 2, 3, 4, and a tilde under a tffiT,C,A,G letter indicates a vector. We get Since reversibility, i.e., ~rtpei(t) = lriPit(t), X3 = -0rv3 + a'Ra), X4 = -(Trva + ~r~) (Sa) can be easily proven, Eq. (7) becomes Pr(x(t) = i, y(t) = j) = fir~ ~ Pi,(t)P,j(t) t ,.~RTgT ~ t V 1 ~ 7rC 71"RTrc / V 2 71"A -- ,R'y'/I'A / = firiP0(2t) 7rG 71"yT/'G/ (Chapman-Kolmogorov equation) Therefore, the average numbers of transition- and transversion-type differences are calculated as fol- V 4 = (5b) lows: V(t) = 2fr~-vlrR[ 1 -- exp(-- 230] (8a) 1/~rv\ S(t) = 2fr{(a'vrc + lrATrG) 1/~rv| "-~ (71"TTrCTrR/Try "21- ~-ATrG~rv/TrR)exp(--2flt) g~ = -1/~r~]' -- 0rTTrc/Try)exp[--2t(aTrv + 31rv3] 1/'lr R ] (5C) -- (TrAlrG/TrR)exp[--2t(a~rR + BTrv)]} (8b)

7l'C/Try~ Furthermore, by using Eqs. (6a), (6b), (la), and (lb), (00 / u,= --~rT/~rv[ variances and covariances among differences are ~:3 = 71.Cr/TrR ' oj calculated as follows: For one pair of sequences, Var(V) = V(1 - V/r) (9a) where Try = 71"T + 7/"C and 71"R = 71"A + 71"G. Var(S) = S(1 - S/r) (90) Now, the numbers of transition-type differences, Cov(V, S) = (9c) SO ~, j2), and transversion-type differences, V(j~, j2), -VS/r between the j ~-th and j:-th sequences are defined as and for two different pairs of sequences (j (l), j2 w) follows: and (j(2), j2r

S(j~, J2) = n..T..C.. + n..C..T.. Cov{S(j nit), j2~ S(j (2), j2(2))} J, J~ J, J2 =r + n.a..~.. + n..G..A_ (6a) {q.. L.. !,'...e,...~;.. gl gl' $2g2 ' jim j2o> j,q~j2(a) J) Ja J, J2 transition V(j,, J2)--'--n..T.A. + n..T..9.. - q.. ~,.. e,...q., e,. ?;..} (9d) jtct)j2o) jtol j2o) + n..c..A.. + n..c..G.. J) J~ J~ J~ Cov{S(jl (,), j2(1)), V(jt (2), j,m)} "~ n..A..T.. "31- n..A..C.. J, J~ J~ J~ =r {q.. t,.. e,,.. g,.. t;.. $1 #I' g2$2' + n..G..T.. + n..o..c.. (6b) transition transversion Jl Ja Jl J2 - q..t,..e,'..q.e,..e;..} (9e) where n..r..c., indicates the number of sites that have j,,,,g(,, L,2>L,~, Jt Ja T in the j ~-th sequence and C in the j2-th sequence Coy{V0(', j2('), VOy >, j2m)) irrespective of other sequences. =r 2; {q.. e,..t,'.. t,.. e;.. Let us consider a pair of sequences separated t j,.'L.) j,,.)g.) million years ago. States of each site of these two gl ~1 ' g2 82' sequences are denoted by x(t) and y(t), respectively. transv~ersion t .... v~ersion Under the assumption of stationarity, we have, for - q.. t,.. e,'..q., t,.. t;.. } (90 i§ j,(')jzo) jna)j212) 165 Least-Squares Fitting of the Data + 1 Var(Sk o)) Since our data consist of two classes of nucleotide sites, all of the variables, all of the parameters except t, and all of the statistics defined above must have Sk(~ -- Sk(ti; fk, ak, Bk) (12) a SUbscript or superscript k to designate the class. We finally reduce the data to the following forms: Since the effects of covariance terms are not negli- gible, we minimize R directly in this paper. In our earlier works (Hasegawa and Yano 1984; Hasegawa 1 7 et al. 1984a, 1985), ni~kt instead of Chjke was used in Vk (I) -- -- ~ Vk(i, j) (10a) 7-i j=_ ~ calculating f~. Since the sample size is small, the covariances estimated in that manner are unstable. 1 7 Sk(i) =--- ~ Sk(i,j), (10b) Therefore, in this work we calculate qUkt iteratively 7-i~_. ~. by means of the Newton method discussed below i=1,2 ..... 6 by setting the values of the parameters as follows: Unless i = j = k = g, q~jkt is given by Where the superscript (i) denotes the i-th splitting in Fig. 1. These values are listed in Table 1. From the central limit theorem, these statistics f ~ ~ 7r.Pxi(2h - tj)P.j(b) Can be regarded as constituting the following vector, * Y 'Pxy(tj--tr.)Pyk(tv.)Py,(tv.) Which follows a multivariate normal distribution: forI

exp{- I(D - I~)'fl-l(D - I~)} Newton Method If we substitute the variance and covariance data To minimize R, the Newton method was carried for R, the approximate maximum likelihood esti- out as follows. If _0 - (t,, t3, t4, ts, t6, t"1, a,, /31, f2, mate of the parameters of the model can be obtained a2,/32)' the iteration algorithm of the Newton meth- by minimizing od is given by R =- (D - I~)'f~-'(D - ~) (11) : a2R ,~-' (aR~ o.+, = O o - \~/,~_. \-~-L~_. (14) la Our earlier works (Hasegawa 1984; Hasegawa et al. 1984a, 1985), as an approximation of this gen- Variances of the Estimates eralized least-squares method, we solved the least- SqUares problem by minimizing From Eq. (11), R is given as a function of D and 0 by

k-l i-1 Var(Vk ~ R(p, 0) = (D - 0(0))'fl-'(D -- 0(0)) where 0(0~) = E[D[ 0]. One can obtain ~(D) by set- "{Vk(i) -- Vk(ti; fk, /3k)}= ting 166

C,d o 3

o 2 o l

o 6 Class 1 sites

CM "t I = 92.27-+ 11.73 Myr (Mouse) _ 2 l = 65 (Bovine) t 2 4 i~3 = 13.30 + 1.54 (Gibbon) = s i~4 = I0.86 -+ 1.24 (Orang) s = 3.67 -+ 0.62 (Gorilla) i~6 = 2.68 -+ 0.61 (Chimp) (D

CD CD c:) T l I I I I l l I I I l I I I I I I I I l I l 0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 V/r

Fig. 2. Least-squares fitting of the relation between S/r and V/r. Vertical and horizontal lines indicate standard deviations of S(~ and VOVr, respectively.The interval between neighboring small circles along the curve is 5 Myr

Var(~) = E[(~ - 0)(~ - 0)'] ~R(p, 6) = 0. (15) "-" B-1AE[(D~ -- ~(0))

Defining (D - I)(0))']A'B-~ = B-JAgA,B-I (17) Rt(D , 0) - ~0 R(D, 0),

we expand R o_(~D, ~ around R e_(I)(~0), ~ as follows: Results R_o(~D, 0~) - Ro(O(O), O) Divergence Times in the Evolution of the A(D - ~(0)) + B(g - 0) (16) Hominoidea where The results are shown in in Figs. 2 and 3. Our clock 0 2 gives 92.27 _+ 11.73, 13.30 _+ 1.54, 10.86 _+ 1.24, A - --R(~(0), 0) 3.67 _+ 0.62, and 2.68 _+ 0.61 Myr for the separatio~ ao oD' from the human line of mouse, gibbon, orangutan, and gorilla, and chimpanzee, respectively (the number after _+ is the standard deviation). (~2 B = __ The estimate of the other parameters of the 30 00/R(~(~0)' 0) model are as follows: t'2 = 0.9491 _+ 0.0395, &l ~ 0.4483 + 0.1424 Myr -~, ~ = 0.0082 + 0.0012 Since the left-hand side of Eq. (16) is zero, MYr-~, f2 = 0.3847 _+ 0.0228, &2 = 0.0684 _+ 0.0093 - 0 = --B-'A(D - 0(~)) Myr -1, ~2 = 0.0062 +- 0.0007 Myr -1 (the number after _+ is the standard deviation). The fact that 31 Therefore, the variance and covariance matrix of nearly equals unity shows that almost all of the third the estimates is given by codon positions are variable. 167

Man

3.rzff o.61 Myr

Chimpanzee Gorilla 10.86 _+ 1.24 Myr/ 13.30 • 1.54 Myr

Orangutan Fig. 3. Estimates of dates of separations during evolution of the Hominoidea. A horizontal line indicates the range of the standard Gibbon error of the estimate

Simulation Experiment function of divergence time t, but has a maximum value and thereafter decreases. This is always the To Confirm the validity of our method, a computer case if a is greater than ~. This characteristic has Simulation was carried out. A hypothetical ancestral not been pointed out by previous researchers, be- sequence corresponding to node 1 in Fig. 1 was con- cause their interest has been in counting the accu- Structed according to the average nucleotide com- mulated number ofnucleotide substitutions, that is, laositions of the respective classes of our data set of transitions plus transversions (Kimura 1980, 1981; rntDNA. In the simulation the sequence evolves Takahata and Kimura 1981; Gojobori et al. 1982). aCCording to the Markov model in which ai, Bi, fi In animal mtDNA, however, in which transition (i ~ 1, 2), and t~ (i = 1..... 6) are the estimates greatly predominates over transversion, this num- Obtained by our analysis to give a set of seven con- ber is not adequate as a measure of genetic distance. ~veraPorary sequences. Then the sequences are ana- Separate counting of transitions and transversions YZed by our method, and the parameters of the is preferable if possible, as is the case for direct model are estimated. sequence data. The simulation was performed 100 times, and The domain of convergence of the Newton meth- the Sample means and sample variances of the es- od is narrow. To find a good initial parameter set, timates were computed. The results are as follows: it is useful to have a good grasp of the characteristics t~ ~ 91.80 _ 11.30 Myr, t3 = 13.24 _ 1.16 Myr, of the relationship between ~r and S, t4 ~ 10.79 __+ 0.88 Myr, t5 = 3.65 + 0.48 Myr, t6 --- 2.73 + 0.43 Myr, fl = 0.9538 ___ 0.0266, al dS a IrTlrc + ~'ATrO 0'4547 -+ 0.1000 Myr -~, fll = 0.0083 __+ 0.0009 MMYr-I, f2 = 0.3852 +_ 0.0181,a2 = 0.0689 _ 0.0076 which is determined by a/~. Furthermore, for yr-1,/~2 = 0.0062 + 0.0006 Myr-1 (the number a># after + is the standard deviation). These results are consistent with the estimates from the DNA se- dS "/I'TTI'CTrR/'/I'y+ "/rATI'G'/I'y/"//"R lim ~ -- quence data, and the standard deviations calculated t-c* dV a'yTrR from the sample variances of simulation experi- .~lents, as well as those calculated from Eq. (17), which is independent of the adjustable parameters. aadieate the degree of error in our estimate of the In Kimura's (1980) formulation, dS/dV tends to Parameters. -1/2 as t goes to infinity. His formula is a good approximation of ours for class 2 sites, where dS/dV 'dOrne Characteristics of the Relationship Between becomes -0.455 as t tends toward infinity, but not and S for class I sites, where dS/dV becomes -0.288. As t goes to infinity, V(t) and S(t) tend to 2frlryTrR t is apparent in Fig. 2 that the number of transition- and 2fr0rTrc + lrA~rG), respectively, which values pe differences S is not a monotonously increasing are dependent on the parameter f. To obtain a good 168 estimate off, distantly related pairs of sequence data, (rl + r2) -1 2rki'k{0rTbrck + 7rAk~rGk)&k in which the number of transition differences de- k=l,2 creases as t increases, are needed. In our data set, + IryklrRk3k} the bovine-primate and mouse-(bovine, primate) pairs are important in this respect. On the other the calculated value of which is (25.4 + 6.1) x 10 -9 hand, to obtain a good estimate of a and B, closely per site per year (the value after _ is the standard related pairs of sequence data, in which the number deviation). This is the average substitution rate of of transition differences increases as t increases, are the segment of 899 nucleotides of mtDNA used in needed. In our data set, the human-chimpanzee and constructing our clock. Although regions outside this gorilla(human, chimpanzee) pairs are important in segment have evolved faster or slower than the seg- this respect. We cannot obtain a good estimate of ment, it seems reasonable to assume that the rate a, B, and f from a single pair of sequences, and estimated above is representative of the average nu- sequence data from many species of organisms, both cleotide substitution rate of the whole mitochon" closely related and distantly related, are needed. Se- drial genome. The above estimate is much larger quence data from intermediately related pairs of than the rates of 2.5 x 10 -9 per site per year, esti- species increase the accuracy of the estimates. The mated by Nei (1982), and 10 x 10 -9 per site per statistical procedure developed in this paper takes year, estimated by Cann et at. (1982). Of the pre- into account the information contained in a set of vious estimates, the rate of 10-20 • 10 -9 per site sequence data more fully than any of the primitive per year, estimated by Brown et al. (1982), is the methods of simple pairwise comparison of se- closest to our value. The divergence times among quences. When mtDNAs from an Old World mon- human races estimated by the previous studies must key and from a New World monkey are sequenced, be revised. The revised divergence times based on the accuracy of our estimate will increase. the molecular clock ofmtDNA are more recent than the estimates obtained from the polymorphism of proteins coded for by nuclear genes (Hasegawa 1984). Average Rate of Nucleotide Substitutions in mtDNA Because mtDNA is more susceptible to transfer between populations, the divergence time estimated Our present study shows that in analyzing direct from the mtDNA clock may indicate the time when sequence data of mtDNA, transitions and transver- mtDNA transfer last happened between two groupS. sions must be treated separately, because the rates of these two kinds of substitutions differ consider- ably. In analyzing restriction endonuclease mapping Discussion data of MtDNA, however, such a separate treatment is impossible. Therefore, it should be useful in in- The Date of Mammalian Divergence terpreting the restriction mapping data to estimate the average rate of nucleotide substitutions in Our datings of the splittings among hominoids are mtDNA irrespective of transition or transversion. heavily dependent on the assumption that the di- Since the previous studies estimated the rate of vergence between bovines and primates occurred 65 nucleotide substitutions in mtDNA without paying Myr ago. Since the holocaust at the end of the Cre- full attention to the fact that a considerable number taceous is likely to have been responsible for the of the transition-type differences in the class 1 sites, mammalian divergence, we think that the date of even those between human and chimpanzee, rep- 65 Myr ago is closer to the truth than previous es- resent multiple hits (Brown et al. 1979, 1982; Nei timates. 1982), their estimates are bound to be lower than Michael Novacek (personal communication) ours. These clocks, which proceed slower than the pointed out that the value of 90 Myr that we adopted real clock, have been used in dating relatively recent for the divergence in the earlier paper (Hasegawa et events; for example, divergences among human races al. 1984a) is unrealistically high, and suggested a have been dated based on restriction endonuclease range between 65 (first appearance of primates and fragment patterns ofmtDNA (Cann et al. 1982; Nei ungulate groups in the fossil record) and 75 Myr ago 1982; Johnson et al. 1983). Therefore, these datings for the split-off from the last common ancestor of must be reexamined by our new molecular clock of primates and bovines. If the older limit of 75 Myr mtDNA. ago is taken, our clock gives 106.46 + 13.54, In a short time interval t, exp(x) in Eqs. (8a) and 15.35 +__ 1.78, 12.53 - 1.43,4.23 _ 0.71,and3.09 (8b) may be approximated by 1 + x. From these 0.71 Myr ago for the separations from the human equations the average rate of nucleotide substitu- line of mouse, gibbon, orangutan, gorilla, and china" tions, that is transitions plus transversions, is there- panzee, respectively. Although there is some uncer" fore given by Hasegawa (1984) tainty as to the date of the mammalian divergence, 169

this divergence event seems to be the most reliable in vain (Colbert 1980), our hypothesis of an earlier reference with which to calibrate the molecular clock rodent divergence may not be unreasonable. of the various references used thus far. In any case, the molecular clock hypothesis as applied to the whole mammalian class is still controversial. It will be desirable to test whether rodent Uniformity of the Rate of the Molecular Clock DNA has evolved more rapidly than that of other placental mammals when an outside reference such The uniformity of the evolutionary rate of mtDNA as marsupial mtDNA is obtained. Even if the mouse among different lineages can be examined by a rel- line has evolved more rapidly than the others, it ative rate test (Wilson et al. 1977). From the data does not invalidate our approach in estimating di- in Table 1, no significant difference is observed vergence times among the Hominoidea. among the numbers of changes between the mouse Although the constancy of our clock with respect and any one of the primates and bovines. Neither to absolute geological time has yet to be proven is any significant difference observed among the directly, future sequencing ofmtDNAs from various numbers of changes between bovines and any one families of primates and from various mammalian of the Primates. Furthermore, no significant differ- orders will clarify the accuracy and applicability of ence is observed among the numbers of changes our clock in estimating divergence times among between the gibbon and other hominoids, and so mammals. It has been proposed that the South on. One might notice that the number of transver- American monkeys descended from African mon- Sion differences observed between gibbon and keys, not from North American prosimians, when orangutan in the class 1 sites, 34, differs consider- the South American continent was close to the Af- ably from the 26 such differences observed between rican continent some 35-38 Myr ago (Ciochon and gibbon and the gorilla--chimpanzee-human trio. ChiareUi 1980). When mtDNA from a new World blowever, this discrepancy is not significant if the monkey is sequenced, the validity of our clock will distribution is Poisson. The number of transition be tested. differences observed between mouse and bovines in the class 1 sites, 39, is also not significantly different from the 46-53 such differences between mouse and Splittings of Orangutan and Gibbon from Primates. Although the possibility of a small devia- Human Line tion from uniformity of the nucleotide substitution l~robability is not excluded, this test shows that our In our earlier paper (Hasegawa et al. 1984a), we data indicate an approximate uniformity at least assumed that the 14.5-Myr-old specimen Sivapith- among primates and bovines. ecus was ancestral to the orangutan but not to hu- Based on nuclear DNA hybridization data, some mans (Raza et al. 1983), and that Micropitheeus, authors have contended that the nucleotide substi- which is some 20 Myr old, was ancestral to the tution rate was much higher along the lineages of gibbons but not to humans (Simons 1981). Assum- mouse and rat than along other mammalian lineages ing 90 Myr ago for the date of the splitting between (Kohne 1970; Kohne et al. 1972). However, their primates and ungulatesand using a least-squares studies were based on questionable estimates of di- method by minimizing R, represented by Eq. (12), Vergence time, as pointed out by Sarich and Wilson we obtained for the divergence dates of the oran- (1973) and by Wilson et al. (1977). Furthermore, gutan and of the gibbons from the human line 8arich and Cronin (1980) and FerNs et al. (1983) 15.9 + 2.9 and 19.1 + 3.6 Myr ago (nijke values suggested that the rates ofnucleotide divergence are instead of q~k e values were used in calculating ~), similar for primates and rodents. Therefore, it is respectively, which are consistent with the above Possible that the rate of mtDNA divergence in ro- interpretation of the fossil evidence. However, when dents does not differ from that for other mammalian we take 65 Myr for the date of mammalian diver- Orders. gence as in this paper, the estimated dates of the At present, we tentatively think that the nucleo- orangutan and gibbon divergences contradict the tide substitution rate of rodents does not differ from above interpretation of the fossil evidence. that of other mammalian orders, and that, as our Peter Andrews (personal communication) point- ~alysis indicates, rodents diverged from other pla- ed out that the 14.5 Myr ago splitting time between cental mammals 92.27 + I 1.73 Myr ago, before the humans and the orangutan (Raza et al. 1983) is ~amrnalian radiation among most of the extant pla- based on the identification of a fragmentary fossil Cental mammalian orders that took place some 65 that is not well dated. Furthermore, he pointed out Myr ago. Since rodents are unique among placental that Micropithecus had not been shown to be an- mammals in that all attempts of paleontologists to cestral to the gibbons. Micropithecus is now recog- relate them to other groups of mammals have been nized as belonging to a primitive group lacking the 170

synapomorphies (derived characters that are shared) of humans and the chimpanzee (and probably the of the Hominoidea (Andrews 1981). Our datings in gorilla). The chimpanzee (and the gorilla) lost bi- this paper, 10.86 + 1.24 and 13.30 ___ 1.54 Myr ago pedalism after splitting from the human line. (2) A. for the orangutan and gibbon divergences, respec- afarensis was not an ancestor of any living primate. tively, are in accord with those of Andrews and Bipedalism evolved independently in at least two Cronin (1982), 10 _ 3 Myr ago for the orangutan lineages, the A. afarensis line and the human line. divergence and 12 + 3 Myr ago for the gibbon di- The first possibility was pointed out by Gribbin vergence, and also approximately in accord with and Cherfas (1982). Although no indication has been those of Sarich and Cronin (1977), 9-11 and 11-13 found that the chimpanzee had a bipedal ancestor, Myr ago for the orangutan and the gibbon diver- this does not necessarily exclude the first possibility. gences, respectively. Zihlman (1979) pointed out that the pygmy chimpanzee Pan paniscus has many morphological features in common with A. africanus, which is be- Possibility That Australopithecus afarensis is Not lieved to have descended from A. afarensis. The an Ancestor of Humans That Evolved After the pygmy chimpanzee is the species most closely re- Human-Ape Splitting lated to the common chimpanzee P. troglodytes among extant hominoids (Zihlman et al. 1978; Fer- There is a widely believed hypothesis that Australo- riset al. 1981a, b; Brown et al. 1982; Sibley and pithecus afarensis, which lived some 3.7 Myr ago at Ahlquist 1984). Furthermore, Feldesman (1982a, b) Laetoli in Tanzania and at Hadar in Ethiopia, is our noticed that A. afarensis clearly resembles P. pan- ancestor and evolved after the human-ape splitting iscus in proximal ulnar morphology. The above ar- (Leakey et al. 1976; Johanson et al. 1978; Johanson guments might be compatible with the first possi- and White 1979; Cronin et al. 1981; White et al. bility, that knuckle-walking of extant chimpanzees 1981). However, our dating of the human-chim- and gorillas evolved from bipedality. panzee splitting by the molecular clock of mtDNA Oxnard (1975, 1984) pointed out that bipedalisna is 2.68 + 0.61 Myr ago, which is more recent that might have evolved not once or even twice, but when A. afarensis lived. Two factors may be taken perhaps several times during the evolution of the into consideration to explain this. First, the dating Hominoidea. Therefore, the second possibility also of the Hadar hominids is still controversial, and they appears likely. He claims that the australopithe" could be younger than 3.4 Myr (Sarna-Wojcicki et cines, including A. afarensis, A. africanus, and A. al. 1985). Second, the divergence between primates robustus, were not structurally closely similar to hu- and ungulates may have happened 75 Myr ago rath- mans, and that they were adapted at least in part to er than 65 Myr ago, and if this is the case, then our an arboreal environment. clock gives 3.09 + 0.71 Myr ago for the human- Although it is likely that the australopithecines chimpanzee separation. Considering these uncer- were capable of bipedalism (but probably in a bio- tainties, the possibility that the hypothesis is com- mechanical mode quite different from that em- patible with our clock cannot be discounted. How- ployed by humans), they might also have been quad- ever, this compatibility rests on fragile ground, and rupedal, especially when climbing in trees. It is now it may be worthwhile to examine the validity of the being recognized that although A. afarensis could hypothesis. walk bipedally, it kept its balance more like a chim- Since A. afarensis walked upright on two legs, panzee does than a human does (Stern and Susman despite the similarity of its brain capacity, dentition, 1981, 1983). Thus, the possibility that A. afarensiS and other features to those of apes, most paleoan- was not an ancestor of humans that evolved after thropologists believe it to be the first hominid. How- the human-ape splitting cannot be ruled out at pres- ever, our molecular clock of mtDNA suggests that ent. the human-ape splitting may have occurred more It is unknown whether the last common ancestor recently than when A. afarensis lived. of human and chimpanzee was like the living chim- Bipedalism is widely believed to have been the panzee or the living human. However, it seems to first step in hominization (Leakey et al. 1976; Day have been widely assumed implicitly that the corn- and Wickens 1980; White 1980), and any fossil pri- mort ancestor of the two species was more like the mates that walked upright have been readily ac- chimpanzee than the human. There has been a ten" cepted as our immediate ancestors. If A. afarensis, dency to view hominid features as specialized and which walked upright, is not an ancestor that evolved those of apes as unspecialized. Any fossil hominoids after the human-ape splitting, as is suggested by our that bear some resemblance to humans have been molecular clock, then the following two explana- readily considered to be human ancestors that tions of the origin of bipedalism are possible: (1)A. evolved after the human-ape splitting. Ramapithe" afarensis, an upright biped, was a common ancestor cines had long been believed to be ancestral to hu- 171

roans based on this type of reasoning. However, they tO another. This may be not only a weakness of are now believed to be ancestors of the living orang- mtDNA analysis, but also a strength, because in utah (Andrews 1982; Andrews and Cronin 1982; such a case the mtDNA clock could provide infor- Pilbeam 1982). mation on the ecological relationship between two Recently Alan Walker pointed out that the orang- species. utan, which had been widely believed to be a highly It has recently been found that mtDNA can pass Specialized ape as compared with the chimpanzee across the species boundary in the mouse (Fen-is et and gorilla, may actually be the living hominoid that al. 1983), in the aquatic frog (Spolsky and Uzzell bears the most extensive resemblance among corn- 1984), and in Drosophila (Powell 1983), and it is temporary descendants to the last common ancestor not impossible that such an event happened among of all the living great apes (Lewin 1983). In this early hominoids. If it did occur, the time derived Sense, the orangutan may be a living fossil. from our clock should reflect it. The sequences of Schwartz (1984) pointed out that humans share nuclear DNA, when they become available, should Uniquely few morphological features with either the make the situation clear. At present the possibility chimpanzee or the gorilla, whereas many features must be considered that our clock reflects a transfer are shared by humans and the orangutan. He con- of mtDNA through hybridization between a proto- eluded that humans and the orangutan may be the human and proto-chimpanzee after the former had closest relatives among the living hominoids. Al- developed bipedalism (Fig. 4). though the molecular evidence shows that his con- When two closely related animal species are geo- clusion is clearly wrong (Goodman 1962, 1963; Sar- graphically contiguous, fertile interspecies hybrids ich and Wilson 1967; Ferris et al. 1981a; Andrews sometimes arise at the boundary zone. Such cases and Cronin 1982; Brown et al. 1982; Hasegawa and are well known in primates, e.g., between anubis Yano 1984; Sibley and Ahlquist 1984), his sugges- baboons (Papio anubis) and hamadryas baboons (P. tion leads us to the following important point. Al- hamadryas) (Nagel 1973; Shotake 1981; Sugawara though brain capacity has increased very much along 1982) and between redtail monkeys (Cercop#hecus the human lineage, not only the orangutan but also ascanius) and blue monkeys (C. mitis) (Aldrich- the human may be living fossils with respect to var- Blake 1968; Macdonald 1984). ious morphological features, whereas the chimpan- Intergroup transfer of males is far more common Zee and the gorilla may have specialized quickly among social primates than that of females, and in after they diverged from the human line. If this is anubis baboons and Japanese and rhesus monkeys, the Case, it is possible that some fossil hominoids which are among the most frequently studied mon- that Were ancestral to the chimpanzee or gorilla but keys, females typically remain in their natal group not the human have been assigned to human ances- (e.g., see Moore 1984). In such a social system, in- tors because of some of their residual features. It is terspecies transfer of mtDNA would seem difficult bow clear that the dating of the branching events even ifinterspecies hybridization occurs frequently. from the fossil record alone is a highly difficult job In contrast, in societies of chimpanzees and gorillas, ~a this circumstance, and that the molecular record intergroup transfer of females and not of males is is USeful for this purpose. routine. Therefore, it appears possible that inter- A theory of human origin must be a theory of species transfer of mtDNA happened between a pro- chimpanzee and gorilla origins too (Zihlman 1979), to-human and a proto-chimpanzee. and to clarify our origin, paleoanthropologists must Ifinterspecies transfer ofmtDNA between proto- Seek not only our ancestors but also fossil creatures human and proto-chimpanzee did indeed occur, it ancestral to the chimpanzee or the gorilla but not is tempting to speculate in which direction the trans- to humans. However, no fossil assigned to be an- fer occurred. The lesser intraspecies polymorphism CeStral only to the chimpanzee or gorilla has yet been of human mtDNA compared with that of chimpan- ~earthed. zee (Ferris et al. 1981b) suggests that the transfer occurred from proto-chimpanzee into proto-human. ~Ssibility of lnterspecies Transfer of Mitochondrial Interspecies transfer of mtDNA has been ob- IVA between Proto-human and Proto-chimpanzee served between the fruit fly Drosophila pseudoob- scura and its sibling species D. persimilis (Powell Since mtDNA is inherited maternally, a molecular 1983). Although male F~ hybrids between the two elgck based on it can give only information on ma- species are sterile, F1 females, the sex that must be ternal lineages. Also, datings of species separation fertile to pass mtDNA, are fertile. Thus even if there from mtDNA data may sometimes be in error be- exists a barrier to interspecies hybridization, the in- catlse of the possibility of introgression of these introgression can occur (Takahata and Slatkin 1984). Clependently segregating organelles from one species Natural interspecies transfer of mtDNA is ob- 172

served from Mus domesticus into M. musculus (Fer- 2.7 +_ O. 6 Myr ago / . Man riset aI. I983), and from Rana lessonae into R. Australopithecus [ ridibunda (Spolsky and Uzzell 1984). The extent of afarensis . ~/f mtDNA divergence between M. domesticus and authentic M. musculus is about 5%, and that between lj//~ Interspecies transfer of R. lessonae and authentic R. ridibunda is about 8%. ~'/1l mitochondrial DNA Since the average substitution rate of mtDNA is estimated to be 0.0254 per site per million years, the interspecies transfer of mtDNA occurred 0.05/(2 x 0.0254) ~ 1 Myr ago in mice and 0.08/(2 x 0.0254) ~ 1.5 Myr in aquatic frogs after the species separated. Therefore, if such an introgression occurred between proto-human and pro- to-chimpanzee 2.68 _+ 0.61 Myr ago, then the human-ape splitting may date back to some 5 Myr ago, as suggested by Sarich and Wilson (1967) and by Sarich and Cronin (1977). Fig. 4. A possible model of interspecies transfer of mtDNA Sibley and Ahlquist (1984) calibrated the molec- between proto-human and proto-chimpanzee ular clock based on their nuclear DNA-DNA hy- ter + is the standard deviation) for the chimpanzee bridization data by assuming that the orangutan di- and gorilla separations, respectively, from the hu- verged from the human line 13-16 Myr ago. Since man line (unpublished data). These datings seem to our molecular clock ofmtDNA gives 10.86 _ 1.24 contradict those from the mtDNA data, and this Myr ago as the date for the orangutan divergence, discrepancy may reflect mtDNA transfer between a we think that the 13 Myr date is nearer the actual proto-human and proto-chimpanzee. value than 16 Myr is, provided mtDNA transfer did The DNA sequence data presently available for not occur between the orangutan and the African setting our molecular clock are limited and the dock apes. If the 13 Myr date is used for the orangutan cannot always determine which one of the various splitting, their data give 6.3 Myr ago as the date for possibilities discussed in this paper is the truth, the human--chimpanzee separation, although the Therefore, future sequencing of DNA, particularly amount of error inherent in their data cannot be nuclear DNA, in conjunction with future fossil find- estimated. ings should throw more light on the origin and evo" Though the relevant data are limited, the hy- lutionary history of our species. In this paper, we pothesis of a late divergence between humans and have demonstrated that chimpanzee and human are apes is also suggested by the amino acid sequence far more closely related genetically than is generally of proteins coded for by nuclear DNA (Goodman believed. A molecular approach can be expected to et at. 1983), and by the rarity of synonymous nu- remain an important tool for elucidating the origin cleotide substitutions in hemoglobin genes between and evolution of mankind. humans and the African apes (Liebhaber and Begley 1983; Scott et al. 1984). After submission of this paper, we learned that Acknowledgments. We are deeply grateful to Professor Alia0 C. Wilson for suggesting the possibility of the recent date of ff~-globin genes, which are pseudogenes of the/~-glo- mammalian divergence adopted in this paper. Thanks are doe bin gene family, from human, chimpanzee, gorilla, to Drs. Peter Andrews and Michael Novacek for helpful sug8es" owl monkey (New World monkey), and lemur (pro- tions, to Professor Joseph Felsenstein for providing us with his simian) have been sequenced (Chang and Slightom computer program for inferring phylogeny and for comments, and to Mr. Hajime Gomi and Mr. Manabu Takano for their help 1984; Goodman et al. 1984; Harris et al. 1984). with the computer programs. We also thank Drs. Hiroshi Mi" Goodman and his coworkers (1984) contended that zutani, Haruhiko Noda, Charles E. Oxnard, David Pilbeam, jacl~ the nucleotide substitution rate was slower along the T. Stem, Naoyuki Takahata, Phillip V. Tobias, and Adrienne L- hominoid lineage than along the New World mon- Zihlman for helpful comments on an early version of the mantt" key lineage. However, the difference is not statisti- script, although their opinions vary greatly. This work was suP" cally significant, and we think that the nucleotide ported by grants from the Ministry of Education, Science, arid Culture of Japan. substitution probability per unit time interval of the xbn-globin gene has been approximately uniform at least among the Anthropoidea. Assuming that the References divergence between the Hominoidea and the New World monkeys occurred 38 Myr ago, our prelim- Aldrich-Blake FPG (1968) A fertile hybrid between two cer" inary analysis of the ~k~-globin data gives dates of copithecus spp. in the Budongo forest, Uganda. Folia prirnat01 5.2 + 0.5 and 5.8 ___ 0.6 Myr ago (the number af- (Basel) 9:15-21 173

Alvarez LW, Alvarez W, Asaro F, Michel HV (1980) Extra- Felsenstein J (1981) Evolutionary trees from DNA sequences: terrestrial cause for the Cretaceous--Tertiary extinction. Sci- a maximum likelihood approach. J Mol Evol 17:368-376 ence 208:1095-1108 Ferris SD, Wilson AC, Brown WM (1981a) Evolutionary tree Alvarez W, Alvarez LW, Asaro F, Michel HV (1984) The end for apes and humans based on cleavage maps of mitochon- of the Cretaceous: Sharp boundary or gradual transition? Sci- drial DNA. Proc Nat[ Acad Sci USA 78:2432-2436 ence 223:1183-1186 Ferris SD, Brown WM, Davidson WS, Wilson AC (19811)) Ex- Anderson S, Bankier AT, Barrell BG, de Bruijn MHL, Coulson tensive polymorphism in the mitochondrial DNA of apes. AR, Drouin J, Eperon IC, Nierlich DP, Roe BA, Sanger F, Proc Natl Acad Sci USA 78:6319-6323 Schreier PH, Smith AJH, Staden R, Young IG (1981) Se- Ferris SD, Sage RD, Huang C-M, Nielsen JT, Ritte U, Wilson quence and organization of the human mitochondrial ge- AC (1983) Flow of mitochondrial DNA across a species norne. Nature 290:457-464 boundary. Proc. Natt. Acad Sci USA 80:2290-2294 Anderson S, de Bruijn MHL, Coulson AR, Eperon IC, Sanger F, Gojobori T, [shii K, Nei M (1982) Estimation of average num- Young IG (1982) The complete sequence of bovine mito- ber of nucleotide substitutions when the rate of substitution chondrial DNA: conserved features of the mammalian mi- varies with nucleotides. J Mol Evol 18:414-423 tOchondrial genome. J Mol Biol 156:683-717 Goodman M (1962) Immunochemistry of the primates and Andrews p (1981) Species diversity and diet in monkeys and primate evolution. Ann NY Acad Sci 102:219-234 apes during the Miocene. In: Stringer CB (ed) Aspects of hu- Goodman M (1963) Serological analysis of the systematics of man evolution. Taylor and Francis, London, pp 25-6 1 recent hominoids. Hum Biol 35:377-436 Andrews p (1982) Hominoid evolution. Nature 295:185-186 Goodman M, Braurtitzer G, Stangl A, Schrank B (1983) Evi- Andrews p, Cronin JE (1982) The relationship of Sivapilhecus dence on human origins from haemoglobins of African apes. and Ramapithecus and the evolution of the orang-utah. Na- Nature 303:546-548 ture 297:541-546 Goodman M, Koop BF, Czelusniak J, Weiss ML, Slightom JL Aquadro CF, GreenbergBD (1983) Human milochondrial DNA (1984) The ~-globingene. Its long evolutionaryhislory ofthe variation and evolution: analysis ofnuedeotide sequences from ~-globin gene family of mammals. J/Viol Biol 180:803-823 seven individuals. Genetics 103:287-312 Gribhin J, Cherfas J (1982) The monkey puzzle. Bodley Head, laarrell BG, Bankier AT, Drouin J (1979) A different genetic London B'" COde in human milochondria. Nature 282:189-194 Harris S, Barrie PA, Weiss ML, Jefereys AJ (1984) The primate lbb MJ, Van Etten RA, Wright CT, Walberg MW, Clayton DA frill gene. An ancient B-globin pseudogene. J Mol Biol 180: (1981) Sequence and gene organization of mouse mitochon- 785-801 drial DNA. Cell 26:167-180 Hasegawa M (1984) The origin and the evolution of man as /~rowa GG, Simpson MV (1982) Novel features of animal inferred from DNA sequences: an introduction to molecular mtDNA evolution as shown by sequences of two rat cyto- anthropology [in Japanese]. Kaimeisha, Tokyo chrome oxidase subunit II genes. Proc Nail Acad Sci USA Hasegawa M~ Yano T (1984) Phylogeny and classification of 79:3246-3250 Hominoidea as inferred from DNA sequence data. Proc Jpn 13ro~ra WM, George M Jr, Wilson AC (1979) Rapid evolution Acad 60B:389-392 of animal mitochndrial DNA. Proe Nail Acad Sci USA 76: Hasegawa M, Yano T, Kishino H (1984a) A new molecular 1967-1971 clock of mitochondrial DNA and the evolution of hominoids. ~rov,,a WM, Prager EM, Wang A, Wilson AC (1982) Mito- Proc Jpn Acad 60B:95-98 choudrial DNA sequences of primates: tempo and mode of Hasegawa M, Yano T, Miyata T (1984b) Evolutionary impli- evolution, j Mol Evol 18:225-239 cations of error amplification in self-replicating and protein C-annRL, Brown WM, Wilson AC (1982) Evolution of human synthesizing machinery. J Mol Evol 20:77-85 naitochondrial DNA: a preliminary report. In: Bonnb-Tamir Hasegawa M, Iida Y, Yano T, Takaiwa F, Iwabuchi M (1985) B, Cohen p, Goodman RN (eds) Human genetics, part A: the Phyogenetic relationships among eukaryotic kingdoms in- Ch Unfolding genome. Alan R Liss, New York, pp 157-165 ferred by ribosmal RNA sequences. J Mol Evol (in press) aug LYE, Slightom JL (1984) Isolation and nucleotide se- Johanson DC, White TD (1979) A systematic assessment of quence analysis of the B-type globin pseudogene from human, early African hominids. Science 202:321-330 C" gorilla and chimpanzee. J Mol Biol 180:767-754 Johanson DC, White TD, Coppens Y (1978) A new species of IOChon RL, Chiarelli AB (1980) Evolutionary biology of the the genus Australopithecus (Primates: Hominidae) from the New World monkeys and continental drift. Plenum Press, Pliocene of eastern Africa. K_irtlandia 28:1-14 New York Johnson MJ, Wallace DC, Ferris SD, Rattazzi MC, Cavaili-Sfor- CiOChon RL, Corruccini RS (1983) New interpretations of ape za LL (1983) Radiation of human mitochondria DNA types and human ancestry. Plenum Press, New York analyzed by restriction endonuclease cleavage patterns. J Mol COlbert EH (1980) Evolution of the vertebrates 3rd ed. John Evol 19:255-271 Wiley & Sons, New York Kimura M (1968) Evolutionary rate at the molecular level. Cronin JE, Boaz NT, Stringer CB, Rak Y (1981) Tempo and Nature 217:624--626 i) raOde in hominid evolution. Nature 292:113-122 Kimura M (1977) Preponderance of synonymous changes as ay MH, Wickens EH (1980) Laetoli Pliocene hominid foot- evidence for the neutral theory of molecular evolution. Nature 3i Prints and bipedalism. Nature 286:385-387 267:275-276 Ckerson RE (1971) The structure of cytochrome c and the KimuraM (1980) A simple method for estimating evolutionary ~' rates of molecular evolution. J Mol Evol 1:26-45 rates of base substitutions through comparative studies of lsenberg JF (1981) The mammalian radiations: an analysis of nucleotide sequences. J Mol Evoi 16:111-120 trends in evolution, adaptation and behavior. University of KimuraM (1981) Estimationofevolutionarydistancesbetween Chicago Press Chicago homologous nucleotide sequences. Proe Nail Acad Sci USA Feldesman MR i1982a) Morphometrics of the ulna of some 78:454-458 /~e Cenozoic "hominoids." Am J Phys Anthropol 57:187 Kimura M (1983) The neutral theory of molecular evolution. ldesman MR (1982b) Morphometric analysis of the distal Cambridge University Press, Cambridge humerus of some Cenozoic catarrhines: the late divergence Kimura M, Ohta T (1974) On some principles governing mo- hypothesis revisited. Am J Phys Anthropol 59:73-95 lecular evolution. Proc Nail Acad Sci USA 71:2848-2852 174

Kohne DE (1970) Evolution of higher-organism DNA. Rev Sama-Wojcicki AM, Meyer CE, Roth PH, Brown FH (1985) Biophys 33:1-48 Ages of tuff beds at East African early hominid sites and KohneDE, ChisconJA, HoyerBH (1972) Evolutionofprimate sediments in the Gulf of Aden. Nature 313:306-308 DNA sequences. J Hum Evol 1:627-644 Savage DE, Russell DE (1983) Mammalian paleofaunas of the Leakey MD, Hay RL, Curtis GH, Drake RE, Jackes MK, White world. Addison-Wesley, London TD (1976) Fossil hominids from the Laetolil Beds. Nature SchwartzJH (1984) The evolutionary relationships ofman and 262:460-466 orang-utans. Nature 308:501-505 Lewin R (1983) Is the orangutan a living fossil? Science 222: Scott AF, Heath P, Trusko S, Boyer SH, Prass W, Gooman M, 1222-1223 Czelusniak J, Chang L-YE, Slightom JL (1984) The se- Liebhaber SA, Begley KA (1983) Structural and evolutionary quence ofthe gorilla fetal globin genes: evidence for multiple analysis of the two chimpanzee a-globin mRNAs. Nucleic gene conversions in human evolution. Mol Biol Evol 1:3714 Acids Res 11:8915-8929 389 Macdonald D (1984) The encyclopaedia of mammals, vol. 1. Shotake T (1981) Population genetical study of natural hybrid- George Allen & Unwin, London, pp 396-397 ization between Papio anubis and P. hamadryas, primates McKenna MC (1975) Toward a phylogenetic classification of 22:285-308 the mammalia. In: Luckett WP, Szalay FS (eds) Phylogeny Sibley CG, Ahlquist JE (1984) The phylogeny of the hominoid of the primates: a multidisciplinary approach Plenum Press, primates, as indicated by DNA-DNA hybridization. J/r New York, pp 21-46 Evol 20:2-15 Miyata T, Hayashida H, Kikuno R, Hasegawa M, Kobayashi M, Simons EL (1976) The fossil record of primate phylogeny. Ill: Koike K (1982) Molecular clock of silent substitution: at Goodman M, Tashian RE (eds) Molecular anthropology. Ple- least six-fold preponderance of silent changes in mitochon- num Press, New York, pp 35-62 drial genes over those in nuclear genes. J Mol Evol 19:28-35 Simons EL (1981) Man's immediate forerunners. Philos TranS Moore J (1984) Female transfer in primates. Int J Primatol R See Lend [Biol] 292:21-41 5:537-589 Spolsky C, Uzzell T (1984) Natural interspecies transfer of Nagel U (1973) A comparison ofanubis baboons, hamadryas mitochondrial DNA in amphibians. Proc Natl Acad Sci uSA baboons and their hybrids at a species border in Ethiopia. 81:5802-5805 Folia Primatol (Basel) 19:104-165 Stern JT Jr, Susman RL (1981) Electromyography of the gluteal Nei M (1982) Evolution of human races at the gene level. In: muscles in Hylobates, Pongo and Pan: implications for the Bonnr-Tamir B, Cohen P, Goodman M (eds) Human ge- evolution of hominid bipedality. Am J Phys Anthropol 55: netics, part A: the unfolding genome. Alan R Liss, New York, 153-166 pp 167-181 Stern JT Jr, Susman RL (1983) The locomotor anatomy of Novacek MJ (1982) Information for molecular studies from Australopithecus afarensis. Am J Phys Anthropo160:279-318 anatomical and fossil evidence on higher eutherian phylogeny. Sugawara K (1982) Sociological comparison between two wild In: Goodman M (ed) Macromolecular sequences in systematic groups of anubis--hamadryas hybrid baboons. Air study and evolutionary biology. Plenum Press, New York, pp 3-41 Monogr 2:73-131 Nozawa K, Shotake T, Kawamoto Y, Tanabe Y (1982) Elec- Takahata N, Kimura M (1981) A model of evolutionary base trophoretically estimated genetic distance and divergence time substitutions and its application with special reference to raP" between chimpanzee and man. Primates 23:432-443 id change of pseudogenes. Genetics 98:641-657 OxnardCE (1975) The place ofthe australopithecines in human Takahata N, Slatkin M (1984) Mitochondrial gene flow. proc evolution: Grounds or doubt? Nature 258:389-395 Natl Acad Sci USA 81:1764-1767 Oxnard CE (1984) The order of man: a biomathematical anat- Templeton AR (1983) Phylogenetic inference from restrictiola omy of the primates. Yale University Press, New Haven endonuclease cleavage site maps with particular reference to Pilbeam D (1982) New hominoid skull material from the Mio- the evolution of humans and the apes. Evolution 37:221-244 cene of Pakistan. Nature 295:232-234 WhiteTD (1980) Evolutionary implications of Pliocene horn" Pilbeam D (1984) The descent of hominoids and hominids. Sci inid footprints. Science 208:175-176 Am 250(3):60--69 White TD, Johanson DC, Kimbel WH (1981) AustralopithecUS PoweU JR (1983) Interspecific cytoplasmic gene flow in the africanus: its phyletic position reconsidered. S Air J Sci 77: absence ofnuciear gene flow: evidence from Drosophila. Prec. 445-470 Natl Acad Sci USA 80:492-495 Wilson AC, Carlson SS, White TJ (1977) Biochemical evolU" Raup DM, Sepkoski JJ Jr (1984) Periodicity of extinctions in tion. Annu Rev Biochem 46:573-639 the geologic past. Prec. Natl Acad Sci USA 81:801-805 Yunis JJ, Prakash O (1982) The origin of man: a chromosomal Raza SM, Barry JC, Pilbeam D, Rose MD, Shah SMI, Ward S pictorial legacy. Science 215:1525-1530 (1983) New hominoid primates from the middle Miocene Zihlman AL (1979) Pygmy chimpanzee morphology and the Chinji Formation, Potwar Plateau, Pakistan. Nature 306:52-54 interpretaton of early hominids. S Afr J Sci 75:165-168 Sarich VM, Cronin JE (1976) Molecular systematics of the Zihlman AL, Cronin JE, Cramer DL, Sarich VM (1978) pygmy primates. In: Goodman M, Tashian RE (eds) Molecular an- chimpanzee as a possible prototype for the common ancestor thropology. Plenum Press, New York, pp 141-170 of humans, chimpanzees and gorillas. Nature 275:744-746 Sarich VM, Cronin JE (1977) Generation length and rates of Zuckerkandl E, Pauling L (1962) Molecular disease, evolutiON, hominoid molecular evolution. Nature 269:354 and genie heterogeneity. In: Kasha M, Pullman B (eds) I4o" Sarich VM, Cronin JE (1980) South American mammal mo- rizons in biochemistry. Academic Press, New York, pp 189- lecular systematics, evolutionary clocks, and continental drift. 225 In: Ciochon RL, Chiarelli AB (eds) Evolutionary biology of Zuckerkandl E, Pauling L (1965) Evolutionary divergence and the New World monkeys and continental drift. Plenum Press, convergence in proteins. In: Bryson V, Vogel HJ (eds) EvolV" New York, pp 399-421 ing genes and proteins. Academic Press, New York, pp 97~ Sarich VM, Wilson AC (1967) Immunological time scale for 166 hominid evolution. Science 158:1200-1203 Sarich VM, Wilson AC (1973) Generation time and genomic evolution in primates. Science 179:1144-1147 Received October 15, 1984/Revised May 14, 1985