GG11CH12-Nei ARI 30 July 2010 20:34 ANNUAL REVIEWS Further The Neutral Theory of Click here for quick links to Annual Reviews content online, including: Molecular Evolution • Other articles in this volume • Top cited articles • Top downloaded articles in the Genomic Era • Our comprehensive search Masatoshi Nei,1 Yoshiyuki Suzuki,1,2 and Masafumi Nozawa1 1Institute of Molecular Evolutionary Genetics and Department of Biology, Pennsylvania State University, University Park, PA 16802; email: [email protected], [email protected] 2Center for Information Biology and DNA Data Bank of Japan, National Institute of Genetics, Mishima, Shizuoka 411-8540, Japan; email: [email protected] Annu. Rev. Genomics Hum. Genet. 2010. Key Words 11:265–89 genetic drift, molecular clock, mutation, natural selection, First published online as a Review in Advance on June 21, 2010 neo-Darwinism, statistical method The Annual Review of Genomics and Human Genetics Abstract is online at genom.annualreviews.org The neutral theory of molecular evolution has been widely accepted This article’s doi: by Universidade Federal de Goias on 01/23/12. For personal use only. 10.1146/annurev-genom-082908-150129 and is the guiding principle for studying evolutionary genomics and the molecular basis of phenotypic evolution. Recent data on genomic Copyright c 2010 by Annual Reviews. All rights reserved evolution are generally consistent with the neutral theory. How- ever, many recently published papers claim the detection of positive 1527-8204/10/0922-0265$20.00 Annu. Rev. Genom. Human Genet. 2010.11:265-289. Downloaded from www.annualreviews.org Darwinian selection via the use of new statistical methods. Examina- tion of these methods has shown that their theoretical bases are not well established and often result in high rates of false-positive and false- negative results. When the deficiencies of these statistical methods are rectified, the results become largely consistent with the neutral theory. At present, genome-wide analyses of natural selection consist of collec- tions of single-locus analyses. However, because phenotypic evolution is controlled by the interaction of many genes, the study of natural selec- tion ought to take such interactions into account. Experimental studies of evolution will also be crucial. 265 GG11CH12-Nei ARI 30 July 2010 20:34 INTRODUCTION The purpose of this review is to evaluate the neutral theory of molecular evolution from The neutral theory of molecular evolution both the theoretical and empirical points of Positive (Darwinian) has been controversial ever since it was pro- view. We are primarily concerned with the gen- selection: natural posed in the 1960s (43, 74, 79). The initial eral features of molecular evolution rather than selection that enhances criticism came primarily from neo-Darwinian thefrequencyofan specific issues. Although we examine recent sta- evolutionists, who were interested in studying allele relative to others tistical tests of the neutral theory critically, our morphological evolution (22, 99, 104, 146). main interest is an evaluation of the legiti- However, as data on molecular evolution macy of the neutral theory to explain molecular accumulated, it became clear that the general and genomic evolution. During the past two pattern of molecular evolution roughly agrees decades, enormous progress has also occurred with that of the neutral theory, though there are in the study of the molecular basis of phenotypic some exceptions (76, 112). Even Mayr (101), evolution and evolution of multigene families, who was a strong selectionist, accepted the which are highly relevant to the neutral theory. concept of neutral evolution at the molecular However, we shall not discuss these issues here level, though he stated that neutral evolution is because they have already been treated in re- of little interest for evolutionists. By contrast, cent review articles and books (e.g., 18, 19, 25, Nei (112, 114) proposed that a substantial 114, 118, 149). portion of morphological evolution is caused by neutral or nearly neutral mutations. In recent years, however, many papers re- DEFINITION OF NEUTRAL porting detection of positive selection have THEORY OF MOLECULAR been published. These papers have been re- EVOLUTION viewed by a number of authors (e.g., 2, 73, 103, Neutral Mutations 121) from the selectionist point of view. Some authors (e.g., 52, 144) have suggested that a In the early 1960s a number of molecular biolo- majority of amino acid substitutions are due to gists sequenced small proteins such as insulins, positive selection and therefore a new theory of cytochrome c, and hemoglobins from various molecular evolution by natural selection should groups of organisms and showed that the ex- be developed. These papers are based on sta- tent of sequence divergence between species in- tistical analyses of genomic data under various creases as the divergence time increases, but the assumptions that are not necessarily satisfied in proteins in different species often have essen- the real world. It is therefore necessary to ex- tially the same function. For example, cow in- amine the validity of the assumptions and the sulin appears functionally equivalent to human by Universidade Federal de Goias on 01/23/12. For personal use only. statistical methods used. It is also important to insulin, such that it could be used as a medica- examine their conclusions from the biological tion for diabetics. Similarly, in vitro experiments point of view. Some authors (e.g., 60, 64, 124) have shown that cytochrome c proteins are ex- have already raised criticisms against the pa- changeable among different mammalian species Annu. Rev. Genom. Human Genet. 2010.11:265-289. Downloaded from www.annualreviews.org pers advocating selectionism. Controversy over as long as the amino acids of the active sites of the neutral theory has a long history, and it is the protein remain the same (69). For these rea- important to know this history to avoid mis- sons, early molecular biologists concluded that understandings of the theory and unnecessary amino acid substitutions outside the active sites arguments. The readers who are not well ac- are mostly neutral or nearly neutral (79). We quainted with the early history are advised to believe this is a “biologically meaningful” defi- read Lewontin (87), Kimura (76), and Nei (112, nition of neutrality. 113). Because of space limitation, we discuss the By contrast, population geneticists have of- history only for a few cases important for our ten defined neutral mutations using mathemati- arguments. cal theory. Fisher (38) and Wright (166) showed 266 Nei · Suzuki · Nozawa GG11CH12-Nei ARI 30 July 2010 20:34 that if the relative fitnesses (Wij) of genotypes generation to generation in nature, and there- = A1A1, A1A2,andA2A2 are given by W11 1, fore this type of definition of neutrality would W = 1 + s ,andW = 1 + 2s , respectively, be meaningless. Note also that when |Ns|=1 12 22 Effective population the probability of fixation (u) of a new mutant the expected time for a new mutation to be fixed size: the number of allele (A2) in the population is in the population (tF) is approximately 4N gen- breeding individuals in 9 − erations (77). This means that if N = 10 and an idealized u = 2s /(1 − e 4Ns)(1) the generation time is 1 year, as in the case of population that has = × 9 genetic diversity where N is the effective population size. Not- some insect species, tF 4 10 years (longer = similar to that of the ing that u for Ns 1 is approximately 50 than the history of life). Even for typical mam- observed population times higher than that for Ns =−1, Fisher malian species, where N may be approximately (38, p. 94) concluded that natural selection is 105 and the generation time is approximately . × 6 very effective, because he believed N is of the 4 years, tF would be 1 6 10 years. How can order of 109. (Fisher did not have the con- such a small s remain constant for the entire cept of effective population size, which is often fixation period? Ohta (127) showed that when much smaller than the actual population size.) s fluctuates with generation and the mean |s¯| In other words, even selection coefficients (s) of |s| is much smaller than the variance, u be- as small as ± 10−9 have significant effects on u. comes practically equal to the probability of fix- For this reason, he became a pan-selectionist, ation of neutral mutations. That is, u is approx- and there was no need for him to examine the imately equal to 1/2N whether s is positive or possibility of neutral mutations. Interestingly, negative. It can be shown that this is true even Kimura (74) used essentially the same definition with N s¯ = 4andN = 105. of neutrality (|2Ns|≤1). However, because A more biologically relevant way of defining he knew the effective population size, he be- neutrality would be to consider a statistically lieved this definition of neutrality would be suf- significant difference between the mean fitness = + ficient (76). Furthermore, the above definition (W¯ M 1 2s ) of the population fixed with the = was a mathematical formality for Kimura, and mutant allele and the mean fitness (W¯ A 1) he was actually interested in neutral or nearly of the population fixed with the ancestral allele. neutral mutations in the biological sense (74). Considering a 30% significance level, Nei (113) Ohta (128–130) distinguished between neutral presented a new√ definition of neutrality, which and nearly neutral mutations in terms of the is given by |s | 2N < 1 approximately. [It value of |2Ns|, but this distinction is not very should be noted that this definition is based on meaningful because she defined nearly neutral the assumption that the progeny size of an indi- mutations as those with 0.2 ≤|2Ns|≤4and vidual follows a Poisson distribution.