Science Wiki

Evolutionary biology and natural selection is the scientific study of the genetic, ecological, and selective mechanisms that drive heritable change and adaptation in biological populations over successive generations. Evolution is the change in the heritable characteristics of biological populations over successive generations. These characteristics are expressed from genes that are passed from parents to offspring during reproduction. Evolution is the unifying framework of the biological sciences. It explains the shared biochemical, genetic, and anatomical homology of living organisms, which traces back to a last universal common ancestor (LUCA), and it explains the diversity of life. Molecular clock analyses place LUCA at roughly 4 billion years ago, with large uncertainty, and the oldest widely accepted evidence of life is at least 3.5 billion years old.

In population genetics, evolution is defined as a change in allele frequencies and haplotype frequencies in the gene pool of a population. In evolutionary biology, quantitative genetics, and evolutionary developmental biology, the definition is extended to changes in the distributions of phenotypes, in developmental trajectories, and in the branching pattern of lineages. The processes involved are mutation, genetic recombination, natural selection, genetic drift, and gene flow. Together they account for adaptation within populations, the origin of new species by speciation, and long-term macroevolutionary trends across geological time.

Darwin and Wallace

Charles Darwin and Alfred Russel Wallace independently proposed natural selection as the mechanism of evolutionary change. Their papers were presented together at the Linnean Society of London in 1858, and Darwin set out the full argument and the evidence in On the Origin of Species (1859). Darwin and Wallace had no mechanism of heredity. Darwin's own theory, pangenesis, was incorrect.

Mendelism, population genetics, and the modern synthesis

The rediscovery of Gregor Mendel's work in 1900 led first to a dispute between Mendelians, who emphasized discrete variation, and biometricians, who studied continuous variation. Ronald Fisher, J. B. S. Haldane, and Sewall Wright reconciled the two during the 1910s to 1930s by showing that continuous traits can result from many Mendelian loci. Their mathematical models of selection, drift, and migration formed the basis of population genetics. Theodosius Dobzhansky, Ernst Mayr, George Gaylord Simpson, Julian Huxley, and G. Ledyard Stebbins connected these models to experimental genetics, systematics, paleontology, and botany in the modern synthesis of the 1930s and 1940s.

Molecular and genomic era

The discovery that DNA is the genetic material (1944 to 1952) and the determination of its structure (1953) made it possible to study evolution at the level of sequences. Motoo Kimura's neutral theory of molecular evolution (1968) and the use of molecular clocks changed the interpretation of molecular data. Carl Woese's comparative analysis of ribosomal RNA led to the three-domain system. Since the 1990s, genomics, phylogenetics, and evolutionary developmental biology have added data on genome structure, gene regulation, and the genetic basis of adaptation. Some biologists argue for an "extended evolutionary synthesis" that gives more weight to developmental bias, phenotypic plasticity, niche construction, and non-genetic inheritance. Others hold that these topics are compatible with the existing framework. The disagreement concerns the relative importance of these factors and not the occurrence of evolution.

Paleontological, Anatomical, and Molecular Evidence

Several independent lines of evidence support common descent and the action of the mechanisms described below:

• Fossil record. Fossils document successive forms and transitional lineages, such as early tetrapods (Tiktaalik), the origin of birds from theropod dinosaurs (Archaeopteryx and feathered non-avian dinosaurs), and the transition from land mammals to whales (Pakicetus and Ambulocetus).
• Comparative anatomy and development. Homologous structures, such as the forelimb skeleton of tetrapods, and vestigial structures show descent with modification from shared ancestors.
• Molecular evidence. Life shares the same basic genetic code and homologous genes for ribosomes and core metabolism. Species share pseudogenes, such as the inactivated GULO gene for vitamin C synthesis in anthropoid primates, and endogenous retroviruses at identical genomic positions. Phylogenies built from independent genes are largely congruent.
• Biogeography. The distribution of species, including the high endemism of oceanic islands and the faunal differences across Wallace's Line, reflects history and dispersal.
• Direct observation. Evolution has been observed in natural populations, such as the beak-size changes of Darwin's finches on Daphne Major studied by Peter and Rosemary Grant, and in laboratory populations. In the E. coli long-term evolution experiment, begun by Richard Lenski in 1988, twelve populations of Escherichia coli have been propagated for more than 75,000 generations. In one population, a variant able to use citrate under aerobic conditions arose after about 31,000 generations. Evolution of antimicrobial resistance and of resistance to pesticides is a further example.

The Hardy-Weinberg principle

The deductive baseline for evolutionary change in Mendelian populations is the Hardy-Weinberg principle, formulated independently in 1908 by G. H. Hardy and Wilhelm Weinberg. It shows that Mendelian segregation and random union of gametes by themselves preserve genetic variation. In the absence of other forces, allele frequencies stay constant, and genotype frequencies reach fixed proportions after one generation of random mating.

Consider a diploid autosomal locus with two alleles, A and a, at frequencies p and q (with p + q = 1). With random union of gametes (panmixia), genotype frequencies in the next generation are:

where p² is the frequency of AA homozygotes, 2_pq_ is the frequency of Aa heterozygotes, and q² is the frequency of aa homozygotes. For a locus with k alleles at frequencies p₁ to p_k, the genotype frequencies are given by the expansion of (Σ_p__i)². For X-linked loci, equilibrium is approached over several generations and not in one.

Evolution occurs when one or more of the assumptions of the model are violated:

1. No mutation. New alleles are not introduced, and alleles do not interconvert.
2. Random mating. Non-random mating changes genotype frequencies without directly changing allele frequencies. If inbreeding has a coefficient F (the inbreeding coefficient of Wright), the genotype frequencies become p² + pqF for AA, 2_pq_(1 − F) for Aa, and q² + pqF for aa. Assortative mating has a similar effect at the loci that determine the trait.
3. No gene flow. Individuals or gametes are not exchanged with other populations.
4. Very large population size. Sampling error in the transmission of gametes (genetic drift) is negligible.
5. No selection and no segregation distortion. All genotypes have the same viability and fecundity, and alleles are transmitted in equal proportions in meiosis, without meiotic drive.

The Price equation

The most general statement of evolutionary change in a population is the Price equation, derived by George R. Price in 1970. It does not depend on the mechanism of inheritance or the ploidy of the organism. Let a parental population have entities indexed by i, each with a character value z_i (which can be an allele frequency, a quantitative trait, or a group property) and an absolute fitness W_i (the number of offspring). Relative fitness is w_i = W_i / W̄, where W̄ is the mean fitness. If the offspring of i have average character value z_i′ = z_i + Δ_z__i, the change in the population mean is exactly:

The first term is the selection term, the covariance between relative fitness and the trait. If the trait is positively associated with reproductive success, selection raises the mean. The second term is the transmission term, the fitness-weighted expected change in the trait between parent and offspring. It includes the effects of mutation bias, meiotic drive, recombination, and changes in the environment of the offspring. Because it is an identity, the Price equation does not by itself predict evolution, but it organizes the causes of change and is used in theories of multilevel selection and kin selection.

Mutation

Mutations are the ultimate source of new genetic variation. They include point substitutions, insertions and deletions (indels), duplications, inversions, translocations, transposon insertions, and changes in chromosome number. In humans, the germline mutation rate is about 1 to 1.5 × 10⁻⁸ per base pair per generation, so each child carries about 60 to 70 new single-nucleotide variants. Rates differ widely among taxa. RNA viruses mutate at about 10⁻⁶ to 10⁻⁴ per nucleotide per replication, whereas DNA-based microbes have genomic mutation rates of about 0.003 per replication (see Drake's rule).

The distribution of fitness effects of new mutations has a large class of deleterious mutations, a class of nearly neutral mutations, and a small class of beneficial mutations. Most new mutations in coding regions that change amino acids are deleterious or nearly neutral. Mutation does not occur in response to need. This was shown by the Luria-Delbrück experiment (1943), in which resistance to bacteriophage in bacteria was found to arise before the selective agent was present.

When deleterious mutations arise repeatedly and are removed by selection, they reach an equilibrium frequency called mutation-selection balance. If μ is the mutation rate to a deleterious allele, s is the selection coefficient against homozygotes, and h is the dominance coefficient, the equilibrium frequency q̂ of the allele is approximately:

This explains why recessive deleterious alleles, including those that cause many genetic diseases, persist in populations at low frequencies.

Natural selection is the differential survival and reproduction of individuals with different heritable traits. It is the only known process that systematically produces adaptation, although not all evolutionary change is adaptive. Following Richard Lewontin, natural selection occurs whenever a population of entities satisfies three conditions:

1. Variation. Individuals differ in morphological, physiological, biochemical, or behavioral traits.
2. Heritability. Some of the variation is passed from parents to offspring, so that offspring resemble their parents more than other individuals.
3. Differential fitness. Individuals with different trait values differ in the number of offspring that they contribute to the next generation. Because resources are limited, populations tend to produce more offspring than can survive, and the variants that are better suited to local conditions are transmitted at higher rates.

Fitness describes the expected contribution of a genotype or phenotype to future generations. It is measured relative to other types in the same population, and it has components of viability, mating success, and fecundity.

Selection at a single locus

Consider a diploid locus with alleles A and a at frequencies p and q. Let the relative fitnesses of AA, Aa, and aa be w_AA = 1, w_Aa = 1 − hs, and w_aa = 1 − s, where s is the selection coefficient against aa and h is the dominance coefficient. The mean fitness is w̄ = p²w_AA + 2_pq_ w_Aa + q²w_aa. The change in the frequency of A in one generation is:

The sign of Δ_p_ is therefore the sign of the slope of mean fitness with respect to allele frequency. With constant genotype fitnesses, selection moves the population uphill on the surface of mean fitness, which is the basis of Sewall Wright's metaphor of the fitness landscape (adaptive landscape). The metaphor is limited, because fitness depends on the frequency of other types in many cases, and because the surfaces for many loci are not simple.

A new beneficial mutation is at risk of being lost by drift while it is rare. For a new additive allele (genotype fitnesses 1, 1 + s, and 1 + 2_s_) in a population of effective size N_e and census size N, Kimura's diffusion approximation gives the probability of fixation of an allele with initial frequency 1/(2_N_) as:

so that for N_e = N and small s the fixation probability is about 2_s_, a result obtained earlier by Haldane. Even a beneficial mutation with s = 0.01 is lost with a probability of about 98 percent.

Fisher's fundamental theorem

Fisher's fundamental theorem of natural selection (1930) states that the rate of increase in mean fitness that is caused by natural selection is equal to the additive genetic variance in fitness. In terms of absolute fitness W:

The theorem is exact for the partial change that is due to selection, and it excludes changes due to mutation, migration, and changes in the environment. Its meaning was disputed for decades. It is often misread as saying that mean fitness always increases.

Quantitative traits and the breeder's equation

For a quantitative trait, the response to selection is given by the breeder's equation:

where R is the change in the mean trait in the next generation, S is the selection differential (the difference in mean between the selected parents and the population), and h² = V_A / V_P is the narrow-sense heritability. Russ Lande generalized it to a set of correlated traits. For a vector of traits z, the change in the mean per generation is:

where G is the additive genetic variance-covariance matrix, P is the phenotypic variance-covariance matrix, s is the vector of selection differentials, and β = P⁻¹s = ∇ ln W̄ is the selection gradient, which measures the direct selection on each trait while holding the other traits constant (the approach of Lande and Arnold). Because the genetic covariances of G (from pleiotropy and linkage disequilibrium) are not zero, selection on one trait causes a correlated response in others, so that the direction of evolution can be deflected from the direction of selection by genetic constraints.

Modes of selection

Selective sweeps, linkage, and background selection

Selection acts on a locus, but nearby neutral variants are affected through linkage. Linkage disequilibrium between loci separated by a recombination fraction r decays at a rate of r per generation:

When a beneficial allele rises to fixation, linked neutral variation is carried with it, which reduces diversity around the site. This is a selective sweep (genetic hitchhiking, described by Maynard Smith and Haigh in 1974). Sweeps can be "hard", when a single new mutation spreads, or "soft", when the beneficial allele is present on several haplotypes. The continual removal of deleterious alleles also reduces diversity at linked sites, an effect called background selection. Linked loci also interfere with each other's selection (the Hill-Robertson effect), which is one proposed advantage of sexual recombination.

Sexual selection

Darwin described sexual selection as selection that arises from differences in mating success and not in survival. It arises from anisogamy and from differences in parental investment, as shown by Robert Trivers, and from the greater variance in reproductive success of the sex that competes for mates, usually males, as in Bateman's principle. It acts in two ways:

• Intrasexual selection is competition among members of one sex for mates, including contests, sperm competition, and mate guarding. It favors weapons, large body size, and related traits.
• Intersexual selection is mate choice, usually by females, for ornaments, songs, or other displays. Models for the evolution of costly ornaments include the Fisherian runaway process (a genetic covariance between the preference and the ornament), the handicap principle of Zahavi, and good-genes models in which ornaments honestly indicate genetic quality or parasite resistance.

Kin selection and inclusive fitness

Altruism, behavior that lowers the lifetime reproduction of the actor and raises that of the recipient, was explained by W. D. Hamilton in 1964 through inclusive fitness and kin selection. A gene for a social behavior can spread if it satisfies Hamilton's rule:

where C is the fitness cost to the actor, B is the fitness benefit to the recipient, and r is the genetic relatedness between them, in the sense of the regression of the recipient's genetic value on the actor's. For diploid full siblings and for parent and offspring, r = 0.5. For half-siblings and grandparent and grandchild, r = 0.25. In haplodiploid insects, full sisters have r = 0.75. Eusociality evolved many times, but haplodiploidy cannot be the sole explanation, because termites are diploid and many haplodiploid species are not eusocial. Ecological factors such as nest defense and the value of monogamy are also important. The relative merits of inclusive fitness and of group-selection models for explaining social evolution are still debated, although the models are mathematically related.

Genetic drift is the random change in allele frequencies caused by the finite sampling of gametes in each generation. It is a stochastic process that operates in addition to selection.

The Wright-Fisher model

In the Wright-Fisher model, a diploid population of effective size N_e has 2_N__e gene copies at a locus, and each generation is formed by sampling 2_N__e copies with replacement from the previous one. The expected allele frequency does not change, but the variance of the change in one generation is:

Drift is therefore weak in large populations and strong in small ones. Without mutation and migration, every allele eventually becomes fixed or lost. The probability that a neutral allele is fixed equals its starting frequency. Expected heterozygosity declines at a rate of 1/(2_N__e) per generation:

Effective population size

The effective population size N_e is the size of an idealized Wright-Fisher population that would show the same rate of drift as the real one. In natural populations it is usually much smaller than the census size N, because of unequal sex ratios, variance in reproductive success, and fluctuations in size. A widely cited meta-analysis found a mean ratio N_e/_N_ of about 0.1 in wild species, although the ratio varies strongly among taxa. With N_m breeding males and N_f breeding females:

For a population whose size varies over t generations, N_e is the harmonic mean of the sizes:

The harmonic mean is dominated by the smallest values, so brief reductions in size have lasting effects on diversity. Two demographic events are especially important:

• A population bottleneck is an acute reduction in size. The survivors carry a sample of the original gene pool. Even after the population recovers, diversity remains low, and rare alleles are lost faster than heterozygosity declines.
• The founder effect occurs when a few individuals colonize a new area. The new population has a sample of the variation in the source, and rare alleles carried by the founders can reach high frequency. Founder effects explain the high frequency of some genetic diseases in isolated human populations.

Coalescent theory

Looking backward in time, the gene copies in a sample are traced to common ancestors by a process called the coalescent. For two gene copies sampled from a diploid population of effective size N_e, the expected time to their common ancestor is 2_N__e generations. If the neutral mutation rate per generation is μ, the expected nucleotide diversity at equilibrium between mutation and drift is:

This relation allows N_e to be estimated from observed diversity. For example, the human long-term N_e estimated from nucleotide diversity (about 10⁻³ per site) and a mutation rate of about 1.2 × 10⁻⁸ is on the order of 10,000, much smaller than the present census size.

Continent-Island Migration and Fixation Index (F_ST)

Gene flow is the movement of alleles between populations by the dispersal of individuals, gametes, pollen, or seeds, followed by successful reproduction. In a continent-island model, an island population with frequency p receives a fraction m of its gene pool each generation from a mainland with frequency p_m:

After t generations, the difference decays as p_t − p_m = (1 − m)^t (p₀ − p_m). Gene flow therefore homogenizes populations and counteracts both drift and local selection. When migration is stronger than selection (roughly m > s), it can prevent local adaptation, a process called migration load or gene swamping.

The balance between drift and migration in a subdivided population is described by Wright's fixation index F_ST, the variance in allele frequency among subpopulations relative to the total expected heterozygosity:

where the approximation is for an island model. If N_em is greater than 1 (more than one effective migrant per generation), F_ST is below about 0.2, and subpopulations are not strongly differentiated by drift. In continuous populations with limited dispersal, genetic differences increase with geographic distance, a pattern called isolation by distance.

Genetic admixture between previously isolated lineages and introgression can transfer adaptive alleles between populations or species. Examples include the transfer of the EPAS1 high-altitude allele from Denisovans to Tibetans, the transfer of wing-pattern alleles among Heliconius butterflies, and about 1 to 2 percent Neanderthal ancestry in present-day non-African humans.

The neutral and nearly neutral theories

The neutral theory of molecular evolution, proposed by Motoo Kimura in 1968 and by Jack Lester King and Thomas Jukes in 1969, holds that most of the genetic variation within species and most differences between species result from the random fixation of selectively neutral or nearly neutral mutations, and not from positive selection. Many new mutations are deleterious and are removed by purifying selection. A neutral mutation arises at a rate μ₀ per gamete per generation, so a diploid population of size N produces 2_N_μ₀ new neutral mutations each generation, and each has a fixation probability of 1/(2_N_). The rate of substitution is therefore:

The substitution rate of neutral mutations equals the mutation rate and is independent of population size. This is the theoretical basis of the molecular clock. The expected sequence divergence between two lineages that separated T generations ago is D = 2_μ_₀T, so divergence times can be estimated when the clock is calibrated against fossils or other dated events. In practice, rates vary among lineages, so relaxed-clock methods are used.

Tomoko Ohta's nearly neutral theory of molecular evolution extends the theory to mutations with small selection coefficients. Their fate depends on the product N_es: if |_s_| is smaller than about 1/(2_N__e), drift dominates. Because N_e differs among species, the same mutation can be effectively neutral in a small population and deleterious in a large one.

Selection on protein-coding genes is inferred by comparing the rate of non-synonymous substitutions per non-synonymous site (d_N) with the rate of synonymous substitutions per synonymous site (d_S). A ratio d_N/_d__S below 1 indicates purifying selection, a ratio near 1 indicates neutral evolution, and a ratio above 1 indicates positive selection. Because the ratio is usually averaged over sites and over time, positive selection is often detected only when it acts on a few sites. Synonymous sites are also not strictly neutral, because of codon usage bias and effects on RNA structure.

Origins of genetic novelty

• Gene duplication. Susumu Ohno proposed in 1970 that duplication is a major source of new genes. Genes duplicate by unequal crossing over, retrotransposition, and whole genome duplication (WGD), including two rounds early in the evolution of vertebrates and an additional round in teleost fishes, and repeated rounds in many flowering plants. Duplicates (paralogs) may be lost by mutations that inactivate them (pseudogenization), may acquire new functions (neofunctionalization), or may divide the ancestral functions between them (subfunctionalization, in the duplication-degeneration-complementation model).
• Exon shuffling, de novo genes, and horizontal gene transfer. Recombination within introns can combine protein domains in new arrangements. Non-coding DNA can acquire open reading frames and become de novo protein-coding genes. In bacteria and archaea, horizontal gene transfer moves whole operons between lineages. Transfer of genes from the endosymbionts that gave rise to mitochondria and plastids to the nucleus was an important part of eukaryogenesis.
• Regulatory evolution. In evolutionary developmental biology, morphological differences between species often involve changes in the regulation of conserved developmental genes, such as Hox genes, Pax6, and the genes of the Wnt, Hedgehog, BMP, Notch, and FGF signaling pathways. Because the coding sequences of these pleiotropic genes are highly constrained, changes in modular _cis_-regulatory elements such as enhancers can alter the place (heterotopy), timing (heterochrony), amount (heterometry), or kind (heterotypy) of expression with fewer pleiotropic effects. The importance of _cis_-regulatory change relative to changes in protein coding sequences is debated, and well-characterized cases of both kinds exist.

Microevolution is the change in allele frequencies within populations. Speciation is the splitting of one lineage into two or more that are reproductively isolated, and it links microevolution to macroevolution.

Species concepts and reproductive isolation

Under Ernst Mayr's biological species concept (1942), species are groups of actually or potentially interbreeding populations that are reproductively isolated from other such groups. The concept is hard to apply to asexual organisms, to fossils, and to groups with frequent hybridization, for which phylogenetic, ecological, and other species concepts are used. Species are often described as separately evolving metapopulation lineages, and the boundaries between them can be gradual.

Reproductive isolation results from barriers that reduce gene flow. They are classified as acting before or after fertilization:

• Prezygotic barriers:
• Ecological or habitat isolation: Populations use different habitats or hosts. An example is the divergence of the fly Rhagoletis pomonella between hawthorn and domesticated apple.
• Temporal isolation: Populations breed at different times of the day, season, or year. An example is the periodical cicadas of the genus Magicicada, with 13-year and 17-year broods.
• Behavioral isolation: Differences in courtship, calls, visual displays, or pheromones prevent mating between species.
• Mechanical isolation: Differences in genitalia or flower structure prevent transfer of gametes or pollen.
• Gametic isolation: Gametes fail to fuse because of divergence in gamete recognition proteins, such as bindin in sea urchins and the IZUMO1 and JUNO proteins in mammals.
• Postzygotic barriers:
• Extrinsic isolation: Hybrids are viable but have an intermediate phenotype that is poorly adapted to the parental niches or that is unattractive to mates.
• Hybrid inviability: Hybrid embryos or young die, for example from developmental incompatibilities or from hybrid necrosis in plants.
• Hybrid sterility: Hybrids are viable but cannot produce functional gametes, for example because chromosomes cannot pair in meiosis. The mule, the offspring of a male donkey (2_n_ = 62) and a female horse (2_n_ = 64), has 63 chromosomes and is almost always sterile.
• Hybrid breakdown: First-generation hybrids are fertile, but later generations have reduced fitness.

The Bateson-Dobzhansky-Muller model and Haldane's rule

Intrinsic postzygotic isolation poses a problem for gradual evolution. At a single locus, evolution from AA to aa would require passing through a low-fitness heterozygote. The Bateson-Dobzhansky-Muller model, developed by William Bateson (1909), Theodosius Dobzhansky (1937), and Hermann Joseph Muller (1942), solves it with incompatibilities between two or more loci:

1. An ancestral population with genotype A₁A₁B₁B₁ is divided into two isolated populations.
2. In population 1, a new allele A₂ becomes fixed at the first locus. It is compatible with B₁.
3. In population 2, a new allele B₂ becomes fixed at the second locus. It is compatible with A₁.
4. When the populations meet, hybrids carry A₂ and B₂ together. These alleles have never been tested together by selection, and their interaction can reduce fitness.

Neither population passes through a fitness valley. Examples of genes that cause hybrid incompatibilities include the Hmr and Lhr genes in Drosophila and Prdm9 in house mice. Orr showed that the number of possible incompatibilities increases at least as fast as the square of the number of substitutions that separate the lineages, an expectation called the snowball effect.

Haldane's rule (1922) states that when one sex is absent, rare, or sterile among the hybrid offspring of two species, it is the heterogametic sex (XY males in mammals and Drosophila, and ZW females in birds and Lepidoptera). The main explanation is the dominance theory. Many incompatibilities are recessive, and the heterogametic sex is hemizygous for the sex chromosome and expresses them. Faster evolution of male-specific traits and conflicts involving meiotic drive also contribute.

Geographic modes of speciation

• Allopatric speciation occurs when populations are separated by a barrier, so that gene flow stops. It is widely considered the most common mode in animals. In vicariant speciation, a range is divided by a barrier such as an uplift, a river, or a sea. An example is the formation of the Isthmus of Panama, which separated "geminate" pairs of marine species such as snapping shrimp of the genus Alpheus about 3 million years ago (the date of closure is debated). In peripatric speciation, a small population at the edge of the range becomes isolated. Models in which founder events trigger rapid change in the whole genome, called founder-flush or genetic revolution models, are controversial, and the evidence for them is limited.
• Parapatric speciation occurs between adjacent populations along an environmental gradient with limited gene flow. Plants on heavy-metal mine tailings next to normal pasture, such as Anthoxanthum odoratum, have been proposed as an example. In a ring species, such as the Ensatina eschscholtzii salamanders around the Central Valley of California and the greenish warbler Phylloscopus trochiloides around the Tibetan Plateau, neighboring populations interbreed, whereas the populations at the ends of the ring are isolated. Hybrid zones can be maintained by a balance of dispersal and selection against hybrids. When hybrids have reduced fitness, selection can favor prezygotic isolation, a process called reinforcement.
• Sympatric speciation occurs without geographic separation. Recombination breaks the association between loci for ecological adaptation and loci for mate choice (Felsenstein's problem). Sympatric speciation by gradual divergence is therefore more likely when one trait influences both ecological adaptation and mating, a "magic trait", and when selection is strong and disruptive. Possible examples include the Midas cichlids of Nicaraguan crater lakes and host races of insects that mate on their host plants. Instant speciation by polyploidy is the best-documented form:
• Autopolyploidy arises when unreduced gametes (2_n_) unite to give a tetraploid. The tetraploid is isolated from its diploid parents because their offspring are triploid and produce unbalanced gametes.
• Allopolyploidy follows hybridization between species, followed by chromosome doubling, which restores pairing and fertility. Bread wheat (Triticum aestivum) is an allohexaploid (2_n_ = 6_x_ = 42) that combines three ancestral diploid genomes (AABBDD). In one estimate, polyploidy accompanied about 15 percent of speciation events in flowering plants and about 31 percent in ferns.

Birth-Death Diversification, Adaptive Radiation, and Extinction

Patterns across geological time reflect the balance between speciation and extinction. For a clade with a per-lineage speciation rate λ and extinction rate μ, the net diversification rate is r_d = λ − μ, and in a simple birth-death model the expected number of lineages grows as:

Real clades show more complex patterns, including slowdowns as ecological niches fill.

• Punctuated equilibrium. Proposed by Niles Eldredge and Stephen Jay Gould in 1972, it holds that many species show long periods of morphological stasis that are interrupted by rapid change that is concentrated around speciation events. It is one of several patterns found in the fossil record, which also contains examples of gradual change.
• Adaptive radiation and key innovations. A lineage that evolves a new trait, a "key innovation" (such as the pharyngeal jaw of cichlids, the amniotic egg, or the flower), or that colonizes an area with empty niches, can diversify rapidly. Examples include the Cambrian explosion, Darwin's finches, Hawaiian silverswords, and the radiation of placental mammals after the end-Cretaceous extinction.
• Coevolution and the Red Queen hypothesis. Interacting species are selective agents for each other. In mutualisms, such as plant and pollinator or mycorrhizal associations, coevolution can favor complementary traits. In antagonistic interactions, such as host and parasite, continuing reciprocal adaptation is needed to maintain fitness. Leigh Van Valen's Red Queen hypothesis (1973) arose from the observation that the probability of extinction of many groups of organisms is roughly constant with the age of the lineage.
• Convergent evolution and deep homology. Unrelated lineages in similar environments often evolve similar traits, such as the streamlined bodies of ichthyosaurs, sharks, and dolphins, or C₄ photosynthesis, which has evolved more than 60 times in flowering plants. Convergent traits often involve the same conserved developmental genes, which is called deep homology.
• Mass extinctions. About 99 percent of all species that have lived are extinct. Most extinction is background extinction. Five major mass extinctions (the end-Ordovician, late Devonian, end-Permian, end-Triassic, and end-Cretaceous) removed large fractions of diversity in short intervals, and the end-Permian event was the most severe. Survival in mass extinctions is not purely random and depends in part on traits such as geographic range, but the rules differ from those of background extinction. Recoveries have been followed by radiations of surviving lineages.

Endosymbiosis, Multicellularity, and Conflict Suppression

Several events in the history of life changed the way that information is stored or the level at which natural selection acts. John Maynard Smith and Eörs Szathmáry called them major transitions. They include the origin of chromosomes, the origin of the eukaryotic cell, the origin of sex, the evolution of multicellular organisms, and the evolution of eusocial societies. In the origin of eukaryotes, an archaeal host cell, probably related to the Asgard archaea, acquired an alphaproteobacterial endosymbiont that became the mitochondrion, and the ancestor of plants and algae acquired a cyanobacterium that became the chloroplast (see endosymbiotic theory). The details of the host cell and the timing of the symbiosis remain under investigation. Each transition required cooperation among units that had been competitors and the evolution of mechanisms that suppress conflict.

Genetic Assimilation and Gene-Culture Coevolution

Some traits are not fixed by genotype. Phenotypic plasticity is the ability of one genotype to produce different phenotypes in different environments. Plasticity can be adaptive, can allow populations to persist in new conditions, and can influence the direction of later genetic change, as in models of genetic assimilation and the Baldwin effect. Niche construction is the modification of the environment by organisms, such as beaver dams or soil modification, which changes the selection on them and on other species. Some epigenetic and cultural information is inherited, and in humans, gene-culture coevolution is well documented, as in the evolution of lactase persistence in pastoralist populations. The relative importance of these processes for the evolution of most traits is a subject of continuing research.

Human activities, including harvesting, agriculture, medicine, urbanization, and climate change, are strong selective forces. The current period of large human influence on the Earth is often called the Anthropocene, an informal term that has not been adopted as a formal geological epoch.

Rates of evolution

Two units are used to compare the rate of phenotypic evolution over different timescales:

1. The darwin, proposed by Haldane in 1949, measures the proportional change in a trait mean per million years:

1. The haldane, proposed by Philip Gingerich, measures the change in the mean in units of pooled phenotypic standard deviations (s_p) per generation g, usually for log-transformed values:

Rates measured over short intervals are much higher than rates measured over long intervals. Contemporary rates in populations that respond to human influence, such as in studies of harvested fish, introduced species, and urban populations, can reach thousands of darwins, whereas long-term rates in the fossil record are usually below one darwin. The relationship with interval length arises partly because directional change reverses and is followed by stasis over long periods, so rates calculated over long intervals underestimate rates of change over short intervals. Examples of human-driven evolution include antibiotic, herbicide, and insecticide resistance, fisheries-induced evolution of earlier maturation and smaller body size in harvested fish such as Atlantic cod (Gadus morhua), although the contributions of genetic change and plasticity are debated, and adaptation of urban populations to heat, pollution, and altered habitats.

Generation time and evolutionary rescue

Species with large populations and short generations, such as microbes, insects, and annual plants, can adapt quickly. Long-lived species with small populations are at greater risk. If the optimum of a trait changes at a constant rate η per unit time, the population mean lags behind the optimum, and the lag imposes a genetic load that reduces mean fitness. Population decline can be reversed if adaptation increases mean fitness to replacement level before the population becomes too small. This is evolutionary rescue. In the classical quantitative genetic model of a moving optimum (Lynch and Lande; Bürger and Lynch), the maximum sustainable rate of environmental change, expressed per generation, is:

where V_A is the additive genetic variance of the trait, r_max is the maximum per-generation growth rate at the optimum, and V_ω = ω² + V_P is the sum of the width of the fitness function and the phenotypic variance, which measures the inverse strength of stabilizing selection. In units of time, the rate is divided by the generation time T_g. The model assumes a constant optimum shift, a stable genetic variance, and a trait under Gaussian stabilizing selection, so its predictions are only approximate.

The threshold is proportional to the genetic variance and inversely proportional to generation time. Habitat fragmentation and overharvesting reduce population size and therefore genetic variance, because V_A tends to increase with N_e. Long-lived vertebrates, reef-building corals, and trees have long generations and so a lower rate of evolution per year. Species that cannot adapt quickly enough, and whose dispersal is too limited to follow shifts in suitable climate (the climate velocity), are at risk of local extinction and of loss of phylogenetic diversity.

Why are most newly arising beneficial mutations lost from a population?

When a novel beneficial allele first arises by mutation in a diploid population of size N, it exists as a single copy at frequency 1/(2_N_). While rare, the allele is subject to strong stochastic sampling error during gamete transmission (genetic drift). As demonstrated by Haldane and Kimura, the fixation probability of a weakly beneficial additive allele with selection coefficient s is approximately 2_s_; thus, an allele conferring a 1% fitness advantage (s = 0.01) still has a ≈98% probability of stochastic loss.

How does the Bateson-Dobzhansky-Muller model explain hybrid sterility without populations crossing a fitness valley?

If postzygotic reproductive isolation required incompatible alleles at a single locus (evolving from AA to aa), an evolving lineage would have to pass through an unfit heterozygous intermediate (Aa). In the Bateson-Dobzhansky-Muller model, an ancestral A₁A₁B₁B₁ population splits into two isolated lineages; one fixes A₂ (compatible with B₁) and the other fixes B₂ (compatible with A₁). Neither population experiences reduced fitness during divergence, yet F₁ hybrids carrying both A₂ and B₂ suffer negative epistatic incompatibility.

What does a d_N/d_S ratio greater than 1 indicate in molecular evolution?

In protein-coding sequences, d_S is the rate of synonymous substitutions per synonymous site (which do not alter the encoded amino acid and approximate the neutral mutation rate), whereas d_N is the rate of non-synonymous substitutions per non-synonymous site. A ratio d_N/_d__S < 1 indicates purifying (negative) selection removing deleterious amino acid changes, d_N/_d__S ≈ 1 reflects neutral drift, and d_N/_d__S > 1 provides statistical evidence of positive Darwinian selection favoring novel amino acid replacements.

• Population genetics
• Natural selection
• Genetic drift
• Speciation
• Molecular evolution
• Phylogenetics
• Evolutionary developmental biology
• Modern synthesis (20th century)