The Evolution of Human Populations: A Molecular Perspective

ley, 1979, 1982). According to this theory, extreme Human evolution exhibits repeated speciations and population bottlenecks would facilitate speciation as conspicuous morphological change: from Australo-well as rapid morphological change. pithecus to Homo habilis, H. erectus, and H. sapiens; The human evolutionary line of descent for the last andfrom their hominoid ancestor to orangutans, goril-four million years (Myr) goes back from Homo sapiens las, chimpanzees, and humans. Theories of founder-to H. erectus, H. habilis, Australopithecus afarensis, event speciation propose that speciation often occurs and A. onamensis. Humans and chimpanzees shared as a consequence of population bottlenecks, down to a last common ancestor about six Myr ago; their last one or very few individual pairs. Proponents of punc-common ancestor with the gorillas lived somewhat ear- tuated equilibrium claim in addition that founder-lier; and the last common ancestor of humans, African event speciation results in rapid morphological apes, and orangutans lived some 15 Myr ago. The suc- change. The major histocompatibility complex (MHC) cession of hominid species from Australopithecus to H. consists of several very polymorphic gene loci. The ge-sapiens occurred in association with some species split- nealogy of 19 human alleles of the DQB1 locus co-ting, as illustrated by such extinct lineages as Australo- alesces more than 30 million years ago, before the di-pithecus africanus, A. robustus, and A. boisei. Earlier vergenceof apes andOld World monkeys. Many human species-splitting events include at least the successive alleles are more closely related to pongid and cerco-divergence of the orangutan, pithecoidalleles than to other human alleles. Using the theory of gene coalescence, we estimate that these polymorphisms require human populations of the der of N (cid:53) 100,000 individuals for the several mil-chimpanzee), lion years. This conclusion is conﬁrmed by computer simulations showing the rate of decay of the polymor-bons microsatellite autosomal

The question arises whether the notable morphologiaddition, that in human evolution no bottlenecks have cal changes that have occurred in hominoid evolution occurred with fewer than several thousand individumay have been triggered by population bottlenecks, als. We evaluate studies of mtDNA, Y-chromosome, and perhaps even extreme bottlenecks down to only one or microsatellite autosomal polymorphisms and cona few pairs of individuals, as postulated by theories of clude that they are consistent with the MHC result that founder-event speciation. The population genetics theno narrow population bottlenecks have occurred in ory of gene coalescence makes it possible to investigate human evolution. The available molecular informathe size of ancestral populations by examining the gene tion favors a recent African origin of modern humans, polymorphisms found in modern populations. The techwho spread out of Africa approximately 100,000 to niques of molecular biology make it feasible to identify 200,000 years ago. © 1996 Academic Press, Inc. these polymorphisms at the DNA level.

The HLA Complex
The process of speciation involves the evolution of reproductive isolation among descendants of individuals The human leucocyte antigene (HLA) complex consists of about 100 genes located in chromosome 6, that previously shared in a common gene pool. The founder-event model of speciation claims that specia-within a DNA segment of four million bp in length. The HLA genes specify molecules with a critical role in tis-tion is often triggered by population bottlenecks (Carson, 1968(Carson, , 1986Mayr, 1963). The theory of punctuated sue compatibility and the defense against foreign substances. These genes are arranged in two distinct equilibrium makes the additional claim that bursts of morphological change occur in association with specia-groups, class I and class II, separated by several score genes that have functions mostly unrelated to the im-tion, whereas stasis prevails over the long time intervals between speciation events (Gould, 1982a,b; Stan-mune response (Fig. 1).

FIG. 1.
Location of some polymorphic genes within the HLA complex in human chromosome 6. There are two sets of genes, class I and class II, separated by a region with unrelated genes. The number of alleles known at a locus is written below the box that indicates the location of the gene. The scale below is in kilobases.
The HLA complex is homologous to the major histo-Age of the DQB1 Human Polymorphism compatibility complex (MHC) of mammals and other Exon 2 of the DQB1 gene consists of 270 nucleotides vertebrates (Kaufman et al., 1995;McDevitt, 1995; that specify β-chain amino acids involved in peptide Schwaiger and Epplen, 1995). MHC molecules present binding. This exon is extremely polymorphic in humans on the surfaces of certain cells bind fragments of proand other hominoids. We have analyzed 64 alleles, of teins (antigens) and present them to thymus-derived which 19 are human, 31 from nonhuman apes, and 14 lymphocytes (T cells) expressing T-cell receptors on from Old World monkeys (Table 1). We have calculated their surfaces. The clone of T lymphocytes that bear the pairwise genetic distance between the 64 alleles usreceptors matching a particular combination of protein ing Kimura's (1983) two-parameter distance. Table 2 is fragment and MHC molecule is stimulated, by the conthe matrix of genetic distances for the 19 HLA alleles. tact with the antigen-presenting cells, to proliferate Figure 2 is a phylogenetic tree of the 64 alleles, conand to initiate the specific arm of the immune response, structed by the neighbor-joining method (Saitou and including the secretion of specific antibodies. The MHC Nei, 1987). molecules thus protect against pathogens and para-It is apparent in Fig. 2 that some human allelic linsites in general.
eages are very old. The clade of 31 alleles on the upper The recognition of protein fragments is mediated by part of the figure includes pongid and hominid alleles, a specialized groove on the surface of the MHC molebut none from cercopithecoids. We may notice that the cule, the so-called peptide-binding region (PBR) comhuman Hs * 0302 allele is more closely related to the posed of some 50 amino acid residues (Bjorkman et al., orangutan Or * 1701than to Hs * 0606 (which in turn is 1987aBrown et al., 1993). The composition of the more closely related to Or * 0603 than to the other two amino acids in the PBR varies from one MHC molecule alleles). If we assume that the divergence of the orangto another, and it is primarily this variation that is reutan and human lineages occurred 15 Myr ago, it folsponsible for the tremendous polymorphism characteristic of the MHC molecules and their encoding genes. In people, as well as in some mammalian species (e.g., the house mouse), scores of alleles may exist at any one  trop, 1994;Bontrop et al., 1995;Klein and Figueroa, Number of 1986;Marsh andBodmer, 1991, 1993;McDevitt, 1995;Code Species alleles O'hUigin et al., 1993;WHO, 1992 (Bergströ m and Gyllenstein, 1995;Fan et al., 1989; Pt Pan troglodytes (chimpanzee) 10 Gyllenstein and Erlich, 1989;Gyllenstein et al., 1990;Gg Gorilla gorilla (gorilla) 10 Lawlor Mayer et al., 1988)

as well as in Or
Pongo pygmaeus (orangutan) 4 Hl Hylobates lar (gibbon) 2 rodents ( Arden and Klein, 1982;Figueroa et al., 1988;McConnell et al., 1988 great apes (Ayala et al., 1994). In the present paper we Ph Papio hamadryas (hamadryas baboon) 2 investigate another class II locus, the DQB1 locus, at which many alleles are known in humans as well as in Note. References for the sequences are: Marsh and Bodmer (1993); O'hUigin et al. (1993); and WHO Nomenclature Committee (1992). other primates. lows that the divergence of Hs * 0302 and Hs * 0606 is Thus, the time of divergence between two such alleles is the same as, or only slightly greater than, the time still older. But the human DQB1 polymorphism is much older than 15 Myr. Toward the lower part of the of the species divergence. The minimum-minimum method attempts to approximate the time of divergence genealogy in Fig. 2, we see that Hs * 0601 is more closely related to several macaque and baboon alleles even more closely by using only the minimum value of all pairwise comparisons between species that diverged (Ma * 0602, Mf * 0602, and * 0601, Mm * 0601, and Ph * 0601) than to the previously mentioned Hs * 0302 at a certain time. The minimum-minimum method may, however, correct excessively, because alleles that and Hs * 0606. This implies that the DQB1 human polymorphism is at least 35 Myr old, if we take this as the have evolved slower than the average will be the ones included in the correlation. In any case, the rates of di-age of the divergence between the apes and the Old World monkeys.
vergence estimated by the two methods are fairly similar, 0.207 and 0.192% per million years, or approxi-The age of the DQB1 human polymorphism may be estimated by calibrating first the rate of evolution of mately a substitution rate of 1 ϫ 10 Ϫ9 per site per year.
We shall use the faster of the two rates, obtained by the DQB1 (exon 2) clock. This has been done in Fig. 3 in two different ways, the ''minimum'' and the ''minimum-the minimum method. Figure 4 is a genealogy of the HLA alleles obtained minimum'' methods (Satta et al., 1991(Satta et al., , 1993. The figure plots the genetic distance between pairs of alleles by the UPGMA method, which assumes constant rates of evolution and thus aligns all 19 alleles at the zero-from different species (ordinate) against the time of divergence between the species (abscissa). In the mini-distance point that corresponds to the present. The genealogy suggests that 8 allele lineages were already in mum method, the rate of evolution is estimated to be the regression on time of the minimum distance ob-existence 15 Myr ago, at the time of the divergence of the orangutan from the lineage of African apes and hu-served for each pair of species that diverged at a certain time. Thus, three values (among all those plotted) are mans; and that 12 allele lineages were in existence 6 Myr ago, at the time of divergence of humans, chimps, used for 15 Myr, namely the minimum distance observed between humans and orangutans, chimps and and gorillas. If we ignore allele 0201, the allelic lineages coalesce around 35 Myr ago. Allele 0201 appears orangs, and gorillas and orangs. In the minimum-minimum method, only the lowest of these three values is to be the most divergent of the HLA alleles in both Figs.
2 and 4, but its remote coalescence in Fig. 4 may be in used; and similarly for the other points along the abscissa. part a distortion due to the imposed condition that all lineages evolve at the same rate. The average genetic The minimum allelic distance between two species is assumed to involve alleles that diverged about the time distance between allele 0201 and all others is 0.134, larger than that observed between any two other alleles, when the species lineages diverged, rather than earlier as would be the case for more differentiated alleles. which is 0.109 between allele 0401 and either 0603 or

Coalescence Theory and the Size of Human Populations
The persistence of 12 DQB1 allelic lineages over the last 6 Myr of hominid evolution implies of necessity that no fewer than six individuals could have existed at any one time throughout that time. In fact, the minimum number of individuals must have been much larger, since the probability that all six individuals in a population are heterozygous, each for two different alleles, is effectively nil. To assess the numbers of individuals required in human ancestral populations in or-  Table 1. The tree is obtained by the we will use the theory of gene coalescence which relates neighbor-joining method ( Saitou and Nei, 1987) based on pairwise the effective number (N) of individuals in a population genetic distances estimated by Kimura's (1983) two-parameter model for those sites sequenced in all 64 alleles. to the coalescence time (T), and thus permits estimating one parameter in terms of the other (T calculated from N, or vice versa). 0604. Using the evolutionary rate of 1 ϫ 10 Ϫ9 substitutions per site per year, the lineage of allele 0201 diverged The coalescence theory examines the genealogical relationships between genes (Griffiths, 1980;Hudson, from the other alleles 67 Myr ago. The DQB1 polymorphism is thought to be older than the DRB1 polymor-1990). According to this theory, all alleles present in an extant pool must have descended from a single allele (to phism and, indeed, one of the oldest among the class II loci (Bergströ m and Gyllenstein, 1995).
which they coalesce). For neutral alleles in a random selection (Takahata, , 1993Takahata and Nei, 1990). In this situation, the theory has the same mathematical structure as in neutral gene genealogy except for a scaling factor, f(s), so that the time to coalescence in generations for a large number of genes is, approximately, If u is the selected mutation rate per gene per generation and s is the selection coefficient, then, approximately, Ϫ3/2 (3)

FIG. 4. Genealogy of 19 DQB1 human alleles obtained with a
Given i sampled genes, the coalescence theory perstandard unweighted pair-group method of averages (UPGMA) algomits one to estimate the number of distinct genes, j, rithm based on the genetic distances given in Table 2. The rate of that existed at a certain time, with time measured in nucleotide divergence, given by I in Fig. 3 smaller for balanced selected alleles than for neutral alleles.
Various sources of evidence suggest that HLA polymating population at equilibrium, the mean coalesmorphisms are subject to balancing selection. One reacence time is given by 4N[1 Ϫ (1/i)] generations, where son to suspect balancing selection is, of course, the longi is the number of sampled genes. For any two genes term persistence of the allelic lineages. The presence of (i ϭ 2), the mean coalescence time reduces to 2N generbalancing selection is manifested by the analysis of the ations; for a large number of genes, the mean coales-DNA sequences of HLA alleles. In codons specifying cence time is, approximately, amino acids of the PBR, variation at the first and second positions is significantly higher than at the third T ϭ 4N, (1) position, and this observation is taken as evidence that positive selection acts on the first two positions (Hedrick, 1994; Nei, 1988, 1992; Kaufman et where T is the number of generations to coalescence. If we assume a long-term generation time of 15 years al., 1995; Klein and O'hUigin, 1994;Potts and Slev, 1995). Moreover, Hill et al. (1991Hill et al. ( , 1992a; see Miller, in the human lineage, an estimate of 35 Myr to coalescence corresponds approximately to 2.3 million genera-1994) have shown that MHC polymorphism may increase resistance to Plasmodium falciparum, the para-tions and N ϭ 575,000 individuals. Two observations are apposite at this point. First is that N as an estimate site responsible for malignant malaria.
Estimates of the magnitude of the selection coeffi-of mean population size has the properties of a harmonic mean. Thus, a mean value of N ϭ 575,000 is com-cient, s, that maintains the MHC polymorphisms vary from locus to locus, but range from 0.0007 to 0.019 patible with much larger population sizes for many generations, but is not compatible with much smaller (Ayala et al., 1994). It seems unlikely that the selection coefficient would be in any case much larger than 0.01-population sizes for very many generations. The second observation is that the estimate in equation (1) has a 0.03, but even larger selection coefficients do not allow for the long-term persistence of polymorphisms except very large variance. Thus, a coalescence determined to have occurred 35 Myr ago is compatible with mean pop-in the presence of large populations (Ayala et al., 1994).
If the heterozygotes' advantage over the homozygotes ulation sizes of, say, N ϭ 250,000 or N ϭ 1,000,000 individuals.
is s ϭ 0.01 and the selected mutation rate per site per gene per generation is u ϭ 2.0 ϫ 10 Ϫ6 , a mean effective Equation (1) applies to selectively neutral alleles (Kingman, 1982a,b;Tajima, 1983; Takahata and Nei, size of approximately N ϭ 131,000 individuals is required for a polymorphism that survives for 3 million 1985; Tavaré, 1984). But the coalescence theory has been extended to allelic genealogies under balancing generations (which amounts to 36-45 Myr if we as-  mutation is expected to appear every one thousand generations in a population of one thousand individuals, which is much too slow to counteract the process of random elimination. In a population of 10,000 individuals, sume a long-term generation of 12-15 years) (Table 3).
however, a new overdominant mutation would appear For s ϭ 0.02 and 0.03, N ϭ 99,000 and 82,500 approxievery 100 generations. Even though a new overdomimately. Persistence of the polymorphism for 2.5 million nant mutation has a high probability of being lost by generations (30-37 Myr) requires, approximately, N ϭ chance in the first or following generations, the process 102,000, 76,000, and 63,000 individuals, respectively, would increase somewhat the number of alleles mainfor s ϭ 0.01, 0.02, and 0.03 (Table 3). The conclusion tained by a population of 10,000 individuals. seems fairly robust, as to the order of magnitude, that the mean effective size of human ancestral populations Population Bottlenecks has been 100,000 individuals throughout hominoid evo- The computer-simulation experiments are consistent lution. A similar conclusion has been derived from the with the conclusion drawn using coalescence theory, analysis of the human DRB1 polymorphism (Ayala, namely that the long-term persistence of the DQB1 al-1995; Ayala et al., 1994;Takahata et al., 1992). lelic lineages requires a mean effective population size Computer Simulation Experiments of the order of 100,000 individuals over the last 35 Myr of hominoid history. When population sizes oscillate, We can test the conclusions obtained in the previous section by means of experiments simulated in a com-small numbers have a disproportionately large effect on the value of N, since the effective population size is puter. The coalescence theory starts with current polymorphisms and makes inferences as to the number of the harmonic mean of the population size over time.
But the question remains whether an occasional popu-generations required for gene coalescence. In the computer simulations the time direction is reversed. The lation bottleneck may have occurred, and how small could the bottleneck be. It has been suggested that a process starts with a certain number of alleles, initially all in identical frequencies, and their fate is ascertained very narrow population bottleneck occurred at the transition from archaic to modern H. sapiens, some as the generations proceed. Each individual, formed by random sampling two genes, reproduces with a proba-100,000 to 200,000 years ago (Cann et al., 1987;Goldstein et al., 1995;Stoneking et al., 1990; Vigilant bility that is s higher for heterozygotes than for homozygotes. The pool of genes collected from the reproduc-et al., 1991; see also Dorit et al., 1995).
The consequences of a bottleneck depend not only on ing individuals is sampled again, and the process repeated generation after generation. The number of the size, N b , of the bottleneck, but also on the number, t b , of bottleneck generations. A useful measure for eval-alleles persisting in the population is scored at 100-generation intervals.
uating the effects of a bottleneck is the ratio N b /t b ,

FIG. 5.
Computer simulation of the loss of alleles over 100,000 generations in a population consisting of (a) 1000 or (b) 10,000 individuals. Initially 20 alleles are present, each at 0.05 frequency. Mating is random and there is no mutation or migration. Heterozygotes reproduce with a probability s higher than homozygotes. Without selection, all alleles but one are soon eliminated. With selection advantages of 0.01 and 0.02, such as operate in real populations, all but two or three alleles are soon eliminated in the population with N ϭ 1,000, but as many as 10-12 are still present after 100,000 generations, when N ϭ 10,000. which if smaller than 10 will have drastic effects in re-tion is weak, such as s ϭ 0.01, N has to be correspondingly large, at least 100, for selection to play a role. The ducing genetic variation (Takahata, 1993). Thus a bottleneck of 100 individuals would substantially reduce persistence of HLA polymorphisms over millions of years requires that the size of human ancestral popula-genetic variation if it would last 10 or more generations. Balancing selection facilitates the persistence of tions be at least Ns ϭ 10 at all times. If s ϭ 0.01, the minimum population size possible at any time would polymorphisms through a bottleneck. But because alleles behave as neutral whenever Ns Ͻ 1, if the selec-be N b ϭ 1000 (Ayala et al., 1994;). The minimum number must have been in fact much larger, several thousand individuals at any time in their evolutionary history. because human population bottlenecks cannot last just a few generations, since many generations are required Adam, Eve, and Other Ancestors for a human population to grow from 1000 to its longterm mean, which we have estimated to be around The evolution of the hominids, from Australopithecus to the emergence of H. erectus, occurred in Africa. The 100,000 individuals. The rate of growth of human populations throughout the Pleistocene has been estimated hominid lineage diverged from the chimpanzee lineage about 6 Myr B.P., and the first known hominid, Ardipi-to be about 0.02% per generation (Spuhler, 1993).
We have used computer simulations to explore the thecus ramidus, lived more than four Myr B.P. (White et al., , 1995WoldeGabriel et al., 1994). H. erectus minimum bottleneck size that would allow the persistence of a certain number of alleles through the bottle-emerged from H. habilis in Africa somewhat before 1.7 Myr B.P., and shortly afterward spread to other conti-neck, allowing for some alleles to be lost through the bottleneck. The results given in Figs. 6 and 7 and (Clark, 1992(Clark, , 1994Clark and Lindly, 1989;Jones et al., 1992). H. erectus fossils from Java have been population to grow back to its long-term population size, the smallest bottleneck that would allow the per-dated 1.81 Ϯ 0.04 and 1.66 Ϯ 0.04 Myr B.P. (Swisher et al., 1994). sistence of 20 allelic lineages, out of 30 present before the bottleneck, with a probability of 95%, is 140-180 Some anthropologists argue that the transition from H. erectus to archaic H. sapiens and later to anatomi-individuals (Fig. 6a). When we take into account the time required for the population to recover to its aver-cally modern humans occurred consonantly in various parts of the Old World. Proponents of this ''multire-age size, the minimum population size at the bottleneck becomes substantially larger. For example, if we as-gional model'' emphasize fossil regional continuity in the transition from H. erectus to archaic and then mod-sume a rate of population increase of R ϭ 1% per generation (which is 50 times greater than the average ern H. sapiens, but postulate that genetic exchange occurred from time to time between populations, so that growth rate of human populations throughout the Pleistocene; Spuhler, 1993, p. 279), a minimum effec-the species evolved as a single gene pool, even though geographic differentiation occurred and persisted, just tive population size of 1080-1150 individuals is required to pass 20 of the preexisting 30 alleles, if the as geographically differentiated populations exist in other animal species (Brä uer, 1992;Clark, 1992; Clark population grows to N ϭ 10,000; and 1180-1280 individuals if the population has to grow to N ϭ 100, 000 and Lindly, 1989;Jones et al., 1992;Waddle, 1994;Wolpoff et al., 1988). (Table 4).
Overdominant selection reduces only slightly the Other scientists argue instead that modern humans first arose in Africa somewhat prior to 100,000 years minimum number of individuals required at the bottleneck. As illustrated in Fig. 6b, when s ϭ 0.02, the B.P. and from there spread throughout the world, replacing elsewhere the preexisting populations of H. bottleneck size for passing 20 of 30 alleles is 130-170 individuals if we ignore the required growth back to erectus or archaic H. sapiens (Goldstein et al., 1995;Horai et al., 1995;Rogers and Jorde, 1995; Ruvolo et long-term numbers, but 980-1030 and 1010-1070 when we take into account population growth to 10,000 al., 1993;Stringer, 1990Stringer, , 1992Stringer and Andrews, 1988). Some proponents of this African replacement and 100,000 individuals, respectively.
If we take only into account the DQB1 genes, it is model argue further that the transition of archaic to modern H. sapiens was associated with a very narrow plausible that a small population bottleneck of about 1000 individuals might have occurred in recent human bottleneck, consisting of only two or very few individuals who are the ancestors of all modern mankind. This evolution. But we cannot ignore that other multipleallele polymorphisms occur in present human popula-proposal is buttressed by an interpretation of mitochondrial DNA analysis showing that the diverse mito-tions, including those shown in Fig. 1 for HLA loci. Thus, for example, 59 human alleles are known at the chondrial DNA sequences found in modern humans coalesce to one ancestral sequence, the ''mitochondrial DRB1 gene locus. The minimum bottleneck required to pass 60 alleles of 70, when we take into account popula-Eve'' or ''mother of us all,'' that existed in Africa about 200,000 years ago (Cann et al., 1987;Stoneking et al., tion growth back to long-term numbers of 100,000, is about 4500 individuals if there is no selection Vigilant et al., 1991). This conclusion has been challenged on grounds concerning (i) whether the co-about 4350 individuals with s ϭ 0.01 (Figs. 7a and 7b). Growth to just 10,000 individuals requires bottlenecks alescence is to Africa, (ii) the time of the coalescence, and (iii) the inference of a population bottleneck (e.g., only slightly smaller. Table 4 summarizes the data obtained for a variety of assumptions. Templeton, 1992). The actual date of coalescence depends on assumptions about evolutionary rates. Based It seems fair to conclude that ancestral human populations are unlikely to have been reduced to fewer than on a time of divergence between humans and chimpan-FIG. 6. Probability that 20 of 30 alleles, initially in identical frequencies, will persist after a bottleneck lasting 10 generations, as a function of population size at the bottleneck, when either (a) alleles are neutral or (b) s ϭ 0.02. The graphs represent averages of 200 computer simulations. A ignores population growth after the bottleneck; B takes into account growth until the population reaches N ϭ 10,000 individuals; C population growth occurs until N ϭ 100,000 individuals. The exponential rate of growth is R ϭ 1.01 per generation, with discrete generations. The 95% value ranges are given in Table 4. zees of 6 Myr, the time to coalescence for mitochondrial Gene genealogies gradually coalesce toward a unique DNA ancestral sequence, whereas individual genealo-DNA polymorphism has been recently estimated at 298,000 years B.P., with a 95% confidence interval of gies increase by a factor of two per generation: an individual has two parents in the previous generation, four 129,000-536,000 years (Ruvolo et al., 1993).
The inference that a narrow bottleneck occurred at ancestors in the generation before that, and so on. (The theoretical number of ancestors for any one individual the time of the coalescence is based on a confusion between gene genealogies and individual genealogies. becomes enormous after some tens of generations, but  Table 4 for 95% value ranges for these and other simulations. ''inbreeding'' occurs: after some generations, ancestors throughout the Pleistocene, 298,000 years to coalescence implies an effective mean population size of 7450 appear more than once in the genealogy.) As we pointed out above, assuming an effective popu-mothers or an effective population size of 14,900 humans. The 95% confidence time estimate of 129,000-lation of N individuals, mean coalescence is 4N generations for neutral nuclear polymorphisms. Mitochon-536,000 years yields 6,450-26,800 generations corresponding to a mean effective population size of 6,450-drial DNA is haploid and maternally inherited; hence, the mean coalescence is 2N f , where N f is the number of 26,800 individuals.
There is one more factor to take into account in the mothers. If we assume 20 years for a human generation Note. The three columns on the right give the minimum number of individuals required for passing 10 (of 20 at the start of the bottleneck), 40 (of 50), and 60 (of 70) alleles, with a 95% probability. In some cases it is assumed that the population grows at a rate R per generation before reaching the equilibrium size N. The selective advantage of heterozygotes is s. Each value is based on 200 computer simulations. calculations just made. We have used mean estimates yields 270,000 years as the expected date for the last common ancestor of modern humans, with 95% confi-to coalescence time, but these estimates have large variances. When the sample of genes is large, the stan-dence limits from 0 to 800,000 years (Dorit et al., 1995).
The Y chromosome is haploid and paternally inher-dard deviation of the mean for nuclear genes is larger than 2N (Nei, 1987, eq. 13.74); for mitochondrial DNA ited. The expected mean coalescence is 2N m , where N m is the number of males. If the human generation is 20 it is larger than N/2. The 95% confidence interval for the number of generations to coalescence will corre-years, the expected coalescence yields an effective population size of 6750 fathers, or 13,500 humans, with a spondingly extend at the upper end to more than 2N (53,600) generations for mitochondrial DNA.
95% confidence upper limit of N ϭ 40,000 individuals. If we take into account the standard deviation of the Thus the mitochondrial DNA sequence data are consistent with a mean effective population size between mean coalescence, the 95% upper limit for N would increase to 80,000 individuals. somewhat less than 10,000 and more than 50,000 individuals throughout the Pleistocene. This population The Y-chromosome results do not provide information about the geographic origin of the ancestral ZFY size is, in turn, consistent with the estimate based on the HLA-DQB1 polymorphism of a long-term mean intron. It would be possible, for example, that the intron might derive from an Australasian ancestor, even population size on the order of 100,000 individuals for human ancestors over the last 30-40 Myr. The mito-if the mtDNA would have been inherited from an African woman. Nor is there any reason to assume that the chondrial DNA data are also consistent with the result that no bottleneck smaller than several thousand indi-ancestral ZFY intron and mtDNA would have lived in the same generation; rather, they might be separated viduals could have happened in hominid history. Dorit et al. (1995) have provided evidence favoring a by tens of thousands of years. It would, of course, be more parsimonious to assume a single ancestral popu-recent replacement of human populations. They have sequenced a 729-bp intron located between the third lation for both the coalescent ZFY intron and the mtDNA, but parsimony has little if any value in this and fourth exons of the Y-chromosome ZFY gene, thought to be involved in the maturation of testes or context.
Evidence favoring African ancestry derives, however, sperm. No sequence variation was observed in a sample of 38 men of diverse geographic origins, representing from the analysis of 30 microsatellite autosomal polymorphisms (Goldstein et al., 1995). The phylogeny de-all the world continents. Comparison with the homologous region of the great apes yields a rate of substitu-rived from average genetic distances separates ancestral African from derived non-African populations. tion of 1.35 ϫ 10 Ϫ9 per site per year. Assuming that substitutions in the intron are neutral, coalescence theory Using a directly measured average mutation rate for lar genetics of speciation and human origins. Proc. Natl. Acad. Sci. microsatellite loci, the deepest split in the phylogeny is USA 91: 6787-6794. dated about 156,000 years ago, which thus estimates Bergström, T., and Gyllenstein, U. (1995) The method of Goldstein et al. (1995) compares Bjorkman, P. J., Saper, M. A., Samraoui, B., Bennett, W. S., Strommulti-locus genetic distances between populations, but inger, J. L., and Wiley, D. C. (1987a). Structure of the human class it does not determine the coalescence of individual I histocompatibility antigen, HLA-A2. Nature (London) 329: 506-512. genes. Therefore, no estimates of population size can be inferred on that basis alone. A significant advantage Bjorkman, P. J., Saper, M. A., Samraoui, B., Bennett, W. S., Strominger, J. L., and Wiley, D. C. (1987b). The foreign antigen binding of the method is precisely that the conclusions are valid site and T cell recognition regions of class I histocompatibility antifor human populations, rather than just for individual gens. Nature (London) 329: 512-521. genes. However, there seems to be no reason to exclude Bontrop, R. E. (1994). Nonhuman primate Mhc- DQA and -DQB sec-the possibility that different genes may have different ond exon nucleotide sequences: A compilation. Immunogenetics 39: populational origins. The average distances would then 81-92. reflect the relative genomic contribution of various an-Bontrop, R. E., Otting, N., Slierendregt, B. L., and Lanchbury, J. S. cestral populations. The results would thus be compati- (1995). Evolution of major histocompatibility complex polymorphisms and T-cell receptor diversity in primates. Immunol. Rev. ble with a model, such as proposed by Clark and others 143: 33-62. (Clark, 1992(Clark, , 1994Clark and Lindly, 1989), in which Bräuer, G. (1992). Africa's place in the evolution of Homo sapiens. In a modern African replacement was concomitant with ''Continuity or Replacement? Controversies in Homo sapiens Evosome regional continuity. An argument in favor of this for the D13S122 locus.