Evolvability is defined as the capacity of a system for adaptive evolution. Evolvability is the ability of a population of organisms to not merely generate genetic diversity, but to generate adaptive genetic diversity, and thereby evolve through natural selection.
In order for a biological organism to evolve by natural selection, there must be a certain minimum probability that new, heritable variants are beneficial. Random mutations, unless they occur in DNA sequences with no function, are expected to be mostly detrimental. Beneficial mutations are always rare, but if they are too rare, then adaptation cannot occur. Early failed efforts to evolve computer programs by random mutation and selection showed that evolvability is not a given, but depends on the representation of the program. Analogously, the evolvability of organisms depends on their genotype-phenotype map. This means that biological genomes are structured in ways that make beneficial changes less unlikely than they would otherwise be. This has been taken as evidence that evolution has created not just fitter organisms, but populations of organisms that are better able to evolve.
Wagner describes two definitions of evolvability. According to the first definition, a biological system is evolvable:
- if its properties show heritable genetic variation, and
- if natural selection can thus change these properties.
According to the second definition, a biological system is evolvable:
- if it can acquire novel functions through genetic change, functions that help the organism survive and reproduce.
For example, consider an enzyme with multiple alleles in the population. Each allele catalyzes the same reaction, but with a different level of activity. However, even after millions of years of evolution, exploring many sequences with similar function, no mutation might exist that gives this enzyme the ability to catalyze a different reaction. Thus, although the enzyme’s activity is evolvable in the first sense, that does not mean that the enzyme's function is evolvable in the second sense. However, every system evolvable in the second sense must also be evolvable in the first.
Pigliucci recognizes three classes of definition, depending on timescale. The first corresponds to Wagner's first, and represents the very short timescales that are described by quantitative genetics. He divides Wagner's second definition into two categories, one representing the intermediate timescales that can be studied using population genetics, and one representing exceedingly rare long-term innovations of form.
Pigliucci's second evolvability definition includes Altenberg's  quantitative concept of evolvability, being not a single number, but the entire upper tail of the fitness distribution of the offspring produced by the population. This quantity was considered a "local" property of the instantaneous state of a population, and its integration over the population's evolutionary trajectory, and over many possible populations, would be necessary to give a more global measure of evolvability.
Generating more variation
More heritable phenotypic variation means more evolvability. While mutation is the ultimate source of heritable variation, its permutations and combinations also make a big difference. Sexual reproduction generates more variation (and thereby evolvability) relative to asexual reproduction (see evolution of sexual reproduction). Evolvability is further increased by generating more variation when an organism is stressed, and thus likely to be less well adapted, but less variation when an organism is doing well. The amount of variation generated can be adjusted in many different ways, for example via the mutation rate, via the probability of sexual vs. asexual reproduction, via the probability of outcrossing vs. inbreeding, via dispersal, and via access to previously cryptic variants through the switching of an evolutionary capacitor.
Enhancement of Selection
Rather than creating more phenotypic variation, some mechanisms increase the intensity and effectiveness with which selection acts on existing phenotypic variation. For example:
- Mating rituals that allow sexual selection on "good genes", and so intensify natural selection
- Large population size
- Recombination decreasing the importance of the Hill-Robertson effect, where different genotypes contain different adaptive mutations. Recombination brings the two alleles together, creating a super-genotype in place of two competing lineages.
- Shorter generation time
Robustness and evolvability
The relationship between robustness and evolvability depends on whether recombination can be ignored. Recombination can generally be ignored in asexual populations and for traits affects by single genes.
Robustness will not increase evolvability in the first sense. In organisms with a high level of robustness, mutations will have smaller phenotypic effects than in organisms with a low level of robustness. Thus, robustness reduces the amount of heritable genetic variation on which selection can act. However, robustness may allow exploration of large regions of genotype space, increasing evolvability according to the second sense.
For example, one reason many proteins are less robust to mutation is that they have marginal thermodynamic stability, and most mutations reduce this stability further. Proteins that are more thermostable can tolerate a wider range of mutations and are more evolvable.
Temporary robustness, or canalisation, may lead to the accumulation of significant quantities of cryptic genetic variation. In a new environment or genetic background, this variation may be revealed and sometimes be adaptive.
Exploration ahead of time
When mutational robustness exists, many mutants will persist in a cryptic state. Mutations tend to fall into two categories, having either a very bad effect or very little effect: few mutations fall somewhere in between. Sometimes, these mutations will not be completely invisible, but still have rare effects, with very low penetrance. When this happens, natural selection weeds out the really bad mutations, while leaving the other mutations relatively unaffected. While evolution has no "foresight" to know which environment will be encountered in the future, some mutations cause major disruption to a basic biological process, and will never be adaptive in any environment. Screening these out in advance leads to preadapted stocks of cryptic genetic variation.
Another way that phenotypes can be explored, prior to strong genetic commitment, is through learning. An organism that learns gets to "sample" several different phenotypes during its early development, and later sticks to whatever worked best. Later in evolution, the optimal phenotype can be genetically assimilated so it becomes the default behavior rather than a rare behavior. This is known as the Baldwin effect, and it can increase evolvability.
Learning biases phenotypes in a beneficial direction. But an exploratory flattening of the fitness landscape can also increase evolvability even when it has no direction, for example when the flattening is a result of random errors in molecular and/or developmental processes. This increase in evolvability can happen when evolution is faced with crossing a "valley" in an adaptive landscape. This means that two mutations exist that are deleterious by themselves, but beneficial in combination. These combinations can evolve more easily when the landscape is first flattened, and the discovered phenotype is then fixed by genetic assimilation.
If every mutation affected every trait, then a mutation that was an improvement for one trait would be a disadvantage for other traits. This means that almost no mutations would be beneficial overall. But if pleiotropy is restricted to within functional modules, then mutations affect only one trait at a time, and adaptation is much less constrained. In a modular gene network, for example, a gene that induces a limited set of other genes that control a specific trait under selection may evolve more readily than one that also induces other gene pathways controlling traits not under selection. Individual genes also exhibit modularity. A mutation in one cis-regulatory element of a gene's promoter region may allow the expression of the gene to be altered only in specific tissues, developmental stages, or environmental conditions rather than changing gene activity in the entire organism simultaneously.
Evolution of evolvability
While variation yielding high evolvability could be useful in the long term, in the short term most of that variation is likely to be a disadvantage. For example, naively it would seem that increasing the mutation rate via a mutator allele would increase evolvability. But as an extreme example, if the mutation rate is too high then all individuals will be dead or at least carry a heavy mutation load. Short-term selection for low variation most of the time is usually thought likely to be more powerful than long-term selection for evolvability, making it difficult for natural selection to cause the evolution of evolvability. Other forces of selection also affect the generation of variation; for example, mutation and recombination may in part be byproducts of mechanisms to cope with DNA damage.
When recombination is low, mutator alleles may still sometimes hitchhike on the success of adaptive mutations that they cause. In this case, selection can take place at the level of the lineage. This may explain why mutators are often seen during experimental evolution of microbes. Mutator alleles can also evolve more easily when they only increase mutation rates in nearby DNA sequences, not across the whole genome: this is known as a contingency locus.
The evolution of evolvability is less controversial if it occurs via the evolution of sexual reproduction, or via the tendency of variation-generating mechanisms to become more active when an organism is stressed. The yeast prion [PSI+] may also be an example of the evolution of evolvability through evolutionary capacitance. An evolutionary capacitor is a switch that turns genetic variation on and off. This is very much like bet-hedging the risk that a future environment will be similar or different. Theoretical models also predict the evolution of evolvability via modularity. When the costs of evolvability are sufficiently short-lived, more evolvable lineages may be the most successful in the long-term. However, the hypothesis that evolvability is an adaptation is often rejected in favor of alternative hypotheses, e.g. minimization of costs.
- Colegrave N, Collins S (May 2008). "Experimental evolution: experimental evolution and evolvability". Heredity 100 (5): 464–70. doi:10.1038/sj.hdy.6801095. PMID 18212804.
- Kirschner M, Gerhart J (1998). "Evolvability". Proceedings of the National Academy of Sciences of the United States of America 95 (15): 8420–8427. doi:10.1073/pnas.95.15.8420. PMC 33871. PMID 9671692.
- Altenberg, Lee (1995). "Genome growth and the evolution of the genotype-phenotype map". Lecture Notes in Computer Science 899: 205–259. doi:10.1007/3-540-59046-3_11.
- Friedberg, R. M. (1958). "A Learning Machine: Part I |". IBM Journal of Research and Development 2 (1): 2–13. doi:10.1147/rd.21.0002.
- Altenberg, Lee (1994). "The evolution of evolvability in genetic programming". In Kinnear, Kenneth. Advances in Genetic Programming: 47–74.
- Wagner GP, Altenberg L (1996). "Complex adaptations and the evolution of evolvability". Evolution 50 (3): 967–976. doi:10.2307/2410639. JSTOR 2410639.
- Wagner A (2005). Robustness and evolvability in living systems. Princeton Studies in Complexity. Princeton University Press. ISBN 0-691-12240-7.
- Pigliucci M (2008). "Is evolvability evolvable?". Nature Reviews Genetics 9 (1): 75–82. doi:10.1038/nrg2278. PMID 18059367.
- Ram, Yoav; Hadany, Lilach (2012). "THE EVOLUTION OF STRESS-INDUCED HYPERMUTATION IN ASEXUAL POPULATIONS". Evolution 66 (7): 2315–2328. doi:10.1111/j.1558-5646.2012.01576.x.
- Olson-Manning, Carrie; MR Wagner; T Mitchell-Olds (December 2012). "Adaptive evolution: Evaluating empirical support for theoretical predictions". Nature Reviews Genetics 13: 867–877. doi:10.1038/nrg3322. PMID 23154809.
- Masel J, Trotter MV (2010). "Robustness and evolvability". Trends in Genetics 26 (9): 406–414. doi:10.1016/j.tig.2010.06.002. PMID 20598394.
- Bloom JD, Labthavikul ST, Otey, CR, Arnold FH (2006). "Protein stability promotes evolvability". Proceedings of the National Academy of Sciences of the United States of America 103 (15): 5869–5874. doi:10.1073/pnas.0510098103. PMC 1458665. PMID 16581913.
- Whitacre and Bender; Bender, Axel (2010). "Degeneracy: a design principle for achieving robustness and evolvability". Journal of Theoretical Biology 263 (1): 143–153. doi:10.1016/j.jtbi.2009.11.008. PMID 19925810. Retrieved 2011-03-11.
- Eyre-Walker A, Keightley, PD (2007). "The distribution of fitness effects of new mutations". Nature Reviews Genetics 8 (8): 610–618. doi:10.1038/nrg2146. PMID 17637733.
- Fudala A, Korona R (2009). "Low frequency of mutations with strongly deleterious but nonlethal fitness effects". Evolution 63 (8): 2164–2171. doi:10.1111/j.1558-5646.2009.00713.x. PMID 19473394.
- Masel, Joanna (March 2006). "Cryptic Genetic Variation Is Enriched for Potential Adaptations". Genetics (Genetics Society of America) 172 (3): 1985–1991. doi:10.1534/genetics.105.051649. PMC 1456269. PMID 16387877.
- Rajon, E., Masel, J. (2011). "Evolution of molecular error rates and the consequences for evolvability". PNAS 108 (3): 1082–1087. doi:10.1073/pnas.1012918108. PMC 3024668. PMID 21199946.
- Hinton GE, Nowlan SJ (1987). "How learning can guide evolution". Complex Systems 1: 495–502.
- Borenstein E, Meilijson I, Ruppin E (2006). "The effect of phenotypic plasticity on evolution in multipeaked fitness landscapes". Journal of Evolutionary Biology 19 (5): 1555–1570. doi:10.1111/j.1420-9101.2006.01125.x. PMID 16910985.
- Kim Y (2007). "Rate of adaptive peak shifts with partial genetic robustness". Evolution 61 (8): 1847–1856. doi:10.1111/j.1558-5646.2007.00166.x. PMID 17683428.
- Whitehead DJ, Wilke CO, Vernazobres D, Bornberg-Bauer E (2008). "The look-ahead effect of phenotypic mutations". Biology Direct 3 (1): 18. doi:10.1186/1745-6150-3-18. PMC 2423361. PMID 18479505.
- Griswold CK, Masel J (2009). "Complex Adaptations Can Drive the Evolution of the Capacitor PSI+, Even with Realistic Rates of Yeast Sex". In Úbeda, Francisco. PLoS Genetics 5 (6): e1000517. doi:10.1371/journal.pgen.1000517. PMC 2686163. PMID 19521499.
- Michod RE (1986). "On fitness and adaptedness and their role in evolutionary explanation". J Hist Biol 19 (2): 289–302. PMID 11611993.
- Eshel I (1973). "Clone-selection and optimal rates of mutation". Journal of Applied Probability 10 (4): 728–738. doi:10.2307/3212376. JSTOR 3212376.
- Masel J, Bergman A, (2003). "The evolution of the evolvability properties of the yeast prion [PSI+]". Evolution 57 (7): 1498–1512. PMID 12940355.
- Lancaster AK, Bardill JP, True HL, Masel J (2010). "The Spontaneous Appearance Rate of the Yeast Prion PSI+ and Its Implications for the Evolution of the Evolvability Properties of the PSI+ System". Genetics 184 (2): 393–400. doi:10.1534/genetics.109.110213. PMC 2828720. PMID 19917766.
- King O, Masel J (2007). "THE EVOLUTION OF BET-HEDGING ADAPTATIONS TO RARE SCENARIOS". Theoretical Population Biology 72 (4): 560–575. doi:10.1016/j.tpb.2007.08.006. PMC 2118055. PMID 17915273.
- Draghi J, Wagner G (2008). "Evolution of evolvability in a developmental model". Theoretical Population Biology 62: 301–315.
- Woods RJ, Barrick JE, Cooper TF, Shrestha U, Kauth MR, Lenski RE (2011). "Second-order selection for evolvability in a large Escherichia coli population". Science 331 (6023): 1433–1436. doi:10.1126/science.1198914. PMC 3176658. PMID 21415350.