Rate, molecular spectrum, and consequencesof human mutationMichael Lynch1Department of Biology, Indiana University, Bloomington, IN 47405This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2009.Contributed by Michael Lynch, December 3, 2009 (sent for review September 13, 2009)Although mutation provides the fuel for phenotypic evolution, italso imposes a substantial burden on fitness through the productionof predominantly deleterious alleles, a matter of concern from ahuman-health perspective. Here, recently established databases onde novo mutations for monogenic disorders are used to estimate therate and molecular spectrum of spontaneously arising mutations andto derive a number of inferences with respect to eukaryotic genomeevolution. Although the human per-generation mutation rate isexceptionally high, on a per-cell division basis, the human germlinemutation rate is lower than that recorded for any other species.Comparison with data from other species demonstrates a universalmutationalbiastoward A/T composition, and leads to the hypothesisthat genome-wide nucleotide composition generally evolves to thepoint at which the power of selection in favor of G/C is approx-imately balanced by the power of random genetic drift, such thatvariation in equilibrium genome-wide nucleotide composition islargely defined by variation in mutation biases. Quantification of thehazards associated with introns reveals that mutations at key splice-site residues are a major source of human mortality. Finally, aconsideration of the long-term consequences of current humanbehavior for deleterious-mutation accumulation leads to the con-clusion that a substantial reduction in human fitness can be expectedover the next few centuries in industrialized societies unless novelmeans of genetic intervention are developed.base substitutions|human genetic disorders|introns|mutation rate|mutational spectrumDespite its central significance to matters of health and phe-notypic evolution, many uncertainties still remain about therate and spectrum of mutations spontaneously arising in thehuman genome (1–3). How frequently do germline and somaticmutations arise, and to what extent does this vary between thesexes? What is the relative incidence of various forms of muta-tions, e.g., missense and nonsense base substitutions, insertions,duplications, and deletions, especially among alterations havingmajor phenotypic effects? How does the mutational spectrum inhumans compare with that in other species? And most impor-tantly, what are the consequences of mutation for the long-termgenetic well-being of our species?In the near future, it should be possible to provide refinedanswers to these and many more questions by sequencing thecomplete genomes of well defined pedigrees and somatic tissues(4, 5). However, it is already possible to achieve a relativelycomplete picture of the point-mutation process from databaseson mutations at loci known to underlie monogenic disorders withmajor phenotypic effects. In the case of autosomal-dominant andX-linked disorders, affected individuals can generally be identi-fied as de novo mutants by comparison with the parental phe-notypes, thereby providing a nearly unbiased view of the rate andspectrum of locus-specific mutations, similar to what has beenachieved with reporter construct studies in microbes (6). Esti-mation of the human mutation rate from the incidence ofmonogenic disorders has a long history, dating back to Haldane(7), but Kondrashov (8) pioneered the use of allelic-sequencesurvey data to gain a broad perspective on the molecularmechanisms of mutation across multiple loci. Since the study ofKondrashov (8), there has been a substantial influx of new dataon more alleles and more disorders, allowing for a refinement ofprior estimates, and the very recent emergence of substantialdata from other species provides a basis for comparative analysis.The material covered herein touches upon a number of issuescentral to our understanding of human genetics and evolution:(i) The analyses provide a revised estimate of the human mutationrate per nucleotide site. (ii) Comparison with data for a diversity ofspecies leads to a hypothesis regarding genome nucleotide com-position that appears to be general across much of cellular life.(iii) An evaluation of the mutational cost of introns demonstratesthat the vulnerability of human genes to degenerative mutations isa function not only of a high mutation rate per nucleotide site butalso of aspects of gene structure. (iv) Taking into consideration therates of mutation at both the germline and somatic-cell levels andtheir likely effects, the consequences of mutation for long-termhuman genetic well-being are explored, and some significant priorconcerns are given credence.ResultsBase-Substitution Mutation Rate. Although an earlier attempt toestimate the human mutation rate using disease-gene data focusedonly on nonsense mutations (8), more accurate estimates of boththe mutational spectrum and per-site rates may be obtained byincluding missense mutations, as done in the present study. Forexample, one limitation of a focus on nonsense mutations is that nomutations to nucleotide C can be detected on the coding strand, asthe three terminationcodons (TAA, TAG, and TGA) are devoid ofC. In addition, because termination codons are A+T rich, andthere is a substantial mutational bias in the direction of A+T(where “+” implies total composition across both strands) pro-duction (9), a focus on these three codons may yield an over-estimate of the overall mutation rate. Finally, the inclusion ofmissense mutations substantially increases sample sizes.For the genes involved in this study, the average rates of base-substitutional mutation are 11.63 (1.80) and 11.22 (3.23) × 10−9persite per generation for autosomal and X-linked loci (SDs inparentheses), respectively (Dataset S1). Modifying the latter esti-mate to scale to a 1:1 incidence of exposure across the sexes yieldsan autosomal equivalent estimate of 14.81 (4.26) × 10−9, which isnot significantly different from the direct autosomal estimate. Anaverage pooled estimate for genes that spend equal time in malesand females is then 12.85 (1.95) × 10−9per site per generation.Many sources of error contribute to the locus-specific estimates inDataset S1, but these will all be subsumed into the SE of the overallmean