Genetic Transilience

Darwin: Variation of Animals and Plants, vol. 2. p. 262. (1868)

28 Quoted by Verlot, 'Des Variétés,' & c., 1865, p. 28.

"The most celebrated horticulturist in France, namely, Vilmorin,28 even maintains that, when any particular variation is desired, the first step is to get the plant to vary in any manner whatever, and to go on selecting the most variable individuals, even though they vary in the wrong direction; for the fixed character of the species being once broken, the desired variation will sooner or later appear."


Genetics. 1980 Apr; 94(4): 1011–1038.
The Theory of Speciation via the Founder Principle
Alan R. Templeton


The founder principle has been used to explain many instances of rapid speciation. Advances from theoretical population genetics are incorporated into Mayr's original founder-effect genetic-revolution model to yield a newer model called the genetic transilience. The basic theoretical edifice lies upon the fact that founder event can sometimes lead to an accumulation of inbreeding and an induction of gametic disequilibrium. This, in turn, causes alleles to be selected more for their homozygous fitness effects and for their effects on a more stable genetic background. Selection occurring in multi-locus systems controlling integrated developmental, physiological, behavioral, etc., traits is particularly sensitive to these founder effects. If sufficient genetic variability exists in the founder population, such multilocus genetic systems can respond to drift and the altered selective forces by undergoing a rapid shift to a new adaptive peak known as the genetic transilience. A genetic transilience is, therefore, most likely to occur when the founder event causes a rapid accumulation of inbreeding without a severe reduction in genetic variability. The implications of this model are then examined for three aspects of the founder-effect genetic-transilience model: the attributes of the ancestral population, the nature of the sampling process used to generate the founders and the attributes of the founder population. The model is used to explain several features of the evolution of the Hawaiian Drosophila, and experimental designs are outlined to test the major predictions of the theory. Hence, this theory of speciation can be tested in the laboratory, using systems and techniques that already exist—a rare attribute of most models of speciation.

The second important effect of inbreeding and homozygosity is that a given allele appears against a narrower spectrum of genetic backgrounds. Thus, due to the founder effect, an allele is selected for its effects on a more limited range of genetic backgrounds. This means that epistatic terms that could not be effectively selected for in the ancestral population can now respond to selection and play a major role in restructuring the fitness properties of the genome. For example, fixation at one critical locus could have cascading fitness effects in a strongly epistatic genetic system.

Unaware of Templeton's work at the time, I wrote the following:

Evolution by Irrelevant Linkages (March 4, 2001)

I've been thinking more (quite a lot) about the notion that selection for one polygenic trait can expose previously hidden genetic and phenotypic variation. This model seems to answer some old questions about evolution — such as "Why do closely allied species differ in several characteristics which have no apparent survival value?" Selectionists (or Darwinists) are obliged to invent separate theories for each of the character differences, but are rarely persuasive. The problem is particularly severe when two or more species maintain their specific differences even when growing side by side and competing successfully.

It is unlikely that Natural Selection would cause or allow one species to "evolve into" a distinct species when and where both the parent species and the child species are equally "fit". Instead we may suppose that the evolution occurred elsewhere, in some more extreme environment where Natural Selection favored a small set of polygenic traits.

If Mendel had been right about independent segregation, evolution would have proceeded very differently than it has. But genes are not perfectly free to segregate, being linked with other genes into chromosomes. Nor are all useful traits unitary. Marquis Wheat, for instance, has broader grains and glumes than ordinary wheat. It was found to differ from older varieties in 6 genes, some with dominant effect, others recessive. No doubt there are yet other genes that would have lesser effects, but a much longer course of selection would be required to break them from their linkages to bring them into useful expression.

Selection for a new polygenic trait — such as adaptation to a new soil type, a change in photoperiod, lengthening or shortening of growing season, and so on — can bring into expression gene-combinations that have not previously been expressed. Not only those currently being favored, but also new combinations of genes (alleles) that are only incidentally linked to the favored genes and combinations.

Once the new but incidental variation has been revealed, it becomes subject to selective pressure. Or not. Irrelevant traits may become fixed because of close linkage to genes that are strongly favored. This is a variation on the "Founder Effect" theory, which allows that some traits that are neither favored nor disfavored may predominate in a population only because they were more common in the founders of that population.

Another problem is the tendency of some species hybrids to revert — more or less quickly — to the parental forms. In these cases the irrelevant differences between species appear to segregate as a single unit, or very few, rather than breaking up to give a perfectly intergrading series of intermediate forms. More linkage!

Initial selection for a new polygenic trait will bring into expression other traits that are irrelevant to the new course of selection. These may be subject to further selection (+ or -), or may be tolerated as "random" variation. Continued selection in the same direction will favor changes in linkage, adding more factors to the polygenic trait, while eliminating other genes (relevant or merely linked). After several to many generations, the favored polygenic trait may become "locked into place" by newly formed linkage groups.

There is usually genetic material available for development of novel polygenic traits. The genes may be distributed throughout the genome, and often neutralized by other genes with contrary effects. That is, + factors (enhancing the trait) will be balanced against - factors (minimizing the trait). We may not assume that + factors are necessarily dominant, or that + and - factors will segregate with perfect freedom. A strong + factor may be linked to a strong - factor (or to several weak - factors), which leads to yet another interesting consequence.

Initial selection for a polygenic trait will favor whatever genes (and their linkage groups) that are readily available. Continued selection in the same direction will favor any breakup of linkages of other + and - factors, which were not freely available at the start of the selection process. Eventually, the genetic nature of the new population may be very different from the first successful specimens. Many genes with small effects may be displaced by fewer genes with larger effects, and with different sets of linked genes/traits.

[This point bears examination. A single, persistent selective pressure may lead to a series of changes in the set of incidentally linked traits. These changes are symptoms of changes in linkage, with little or no direct role in the evolutionary process.]

A dozen + genes may be replaced over time by a few +++ genes that were previously masked by — and —- genes linked to them. Thus a single selective pressure may act quickly by favoring freely available + factors, then lock the population into the new form with a smaller set of +++ factors, as they become available. If the modified "species" then has opportunity to backcross to its parent, or other relative, the original set of "polygenes" (each with small effect) may be bred out, leaving only the few "oligogenes" (each having large effect). The new species is now distinguished from the parent by a few or several distinctive traits, few of which may have any apparent relevance to its specific adaptations.

In this hypothetical case, evolution climbs up a polygenic ladder, but becomes isolated on an oligogenic platform. The rungs of the ladder (polygenes) will easily be lost in backcrosses to the parent or other relatives, leaving only the oligogenes and whatever genes/traits are tightly linked to them. In fact, the original polygenes that built up the trait in the first place may later appear to be only "modifying genes".

It is important to remember that while selective pressure is working powerfully in one direction, innumerable physiological processes must also be maintained. The population may become adapted to extremely low pH, but if this adaption interfered with the plants' ability to resist infection, or to reproduce, the population would be doomed to extinction. The modified population might still survive in a neutral soil, but we can make no definite predictions about how it might react in more alkaline areas. Since there has been no recent selection for response to high pH, we may expect more inherent variation in response to high pH. This has another interesting consequence.

If we cross a low pH adapted form with a high pH relative, the appearance of the progeny and the patterns of segregation may vary according to pH, not according to Mendel. All the genes, gene combinations, tissues and traits of the low pH species are stable at low pH, but uncertain (untested) at high pH. Likewise, the genes and traits of the high pH species are stable and true-breeding in high pH conditions, but uncertain at low.

Under low pH conditions, the genes of the low pH species will tend to predominate in the F1 hybrids, and often appear dominant in later generations. But when the same F1 hybrids are raised in high pH conditions, the traits of the other parent will tend to predominate. Free Mendelian segregation will occur most frequently where pH is neutral, or where there is a mingling of low and high pH.

These are not proper predictions, by the way, since the phenomena discussed have already been described (Ivan Michurin, Edgar Anderson, O. F. Cook, G. Turesson and others). Even so, they do follow as consequences from the hypothesis of Irrelevant Linkage.

Another useful consequence answers an objection to a theory of evolution by mutations. If gene mutations occur at random (relative to the needs of the species), they must be produced most frequently when the population is large. But a large population is presumably thriving, and in no great need of new variation. Conversely, few mutations may be expected in a small population on the brink of extinction, and in desperate need of fresh variation.

Harmful traits are commonly more harmful when an organism is stressed, or struggling in an extreme habitat. If its effects are recessive, or nearly so, even a lethal mutation may be preserved in a congenial environment where outcrossing is the rule. Thus, a large and thriving population not only produces more mutations than one which is small and struggling, it also is more able to absorb, disperse and otherwise conceal new mutations, as well as hidden genes of indeterminate age.

Paradoxically, a large and thriving population may conceal extensive genetic variation under a mask of uniformity, while a smaller population in extreme conditions may exhibit great variation though possessing less genetic diversity than the larger group. I mentioned this once before in a different context; that a reduction in genetic diversity can lead to an increase in visible variation. Inbreeding, with or without selection, may convert potential variation (hidden genetic diversity) into actual [phenotypic] variation.

The present treatment of the above ideas and consequences is my own, but I owe much to several earlier writers, and especially to:

The fundamental aspects of genetic transilience are polymorphism, epistasis and linkage.

1) Polymorphism. Within a species or population, a gene may exist in two or more distinct forms (alleles). Mendel's work with peas dealt with alternative unit characters. In most of the cases, the differences were due to distinct forms of a single gene. E.g., Peas normally have purple flowers. White-flowered forms are due to a mutation that blocked synthesis of anthocyanin.

Polymorphism may be cryptic, as in human blood types, and the different forms of esterase studied by Schwartz (1960) in maize.

2) Epistasis. Many traits involve the combined actions of alleles of two or more genes. E.e., Philiptschenko found that the remarkably broad grains and glumes of the 'Marquis' wheat involved alleles of six genes. The genetic formula is [aa + (BB + CC) + (dd + ee + ff)].

Hurst (1932) discussed Philiptschenko's work along with other other examples.

Clausen (1930) described an even more complicated model for flower color in Viola Wittrockiana Gams. that involves alleles of 9 genes.

Austin (1961) found various modifiers of the "horned" trait in tall bearded irises. 'Happy Birthday', for example, showed no trace of the trait, but contributed a gene that enhanced the trait.

One interesting consequence of epistasis is that with the proper arrangement of modifiers, one of the pairs of alleles may segregate as "unit characters". In the case of 'Marquis' wheat, the aa pair organizes the broad grains and glumes, but the trait is not fully expressed without the other genes. However, a plant inheriting Aa or AA would could not be 'Marquis' regardless of the other gene/alleles present. Thus, a cross between a 'Marquis' wheat [aa + (BB + CC) + (dd + ee + ff)] and a derived line with the genetic formula [AA + (BB + CC) + (dd + ee + ff)] would result in a non-Marquis F1. The F2 generation would segregate for 1 Marquis to 3 non-Marquis because the same modifiers are derived from both parents. This creates the illusion that the Marquis trait rests on a single gene.

3) Linkage. Dunn and Caspari (1945), "... gene loci are scattered through the chromosomes without regard to developmental effect. Loci which are so close as to be seldom separated by crossing over do not necessarily resemble each other in their effects, while those with chief effects on very different structures or functions may occur near together or far apart." Anderson (1939a), Anderson (1939b),

Genetic transilience rests on the fact that selection for a multifactoral trait will also bring together linked genes (alleles) in combinations that would be exceedling rare where the particular selective pressure does not occur. This suggests that selection for a genuinely novel compound trait is likely to expose unsuspected traits.

The new traits may be useful. If so, further selection would tend to enhance these traits, and also bring out yet other hidden treasures. On the down side, it is also possible that some new traits are highly undesirable, such as the loss of fragrance in sweet peas that followed (directly?) selection for larger and more numerous flowers on longer, stronger stems.

Theory is fine, but a practical example is helpful. Carrière (1869) provided just such an example with the wild radish, Raphanus raphanistrum.

Having stated what are the principal characters of R. raphanistrum, I will mention how the idea occurred to me of subjecting the plant to cultivation. Born and brought up in the country, most of my youth was spent in the fields. One day I observed some resemblance between the pods of the garden Radish and those of the wild one, which at that time I mistook for Charlock. I found them good to eat, and it occurred to me to sow the seeds, but my intention to do so was not carried out till long afterwards, when, hearing of various experiments made with Cabbages, Beet, and Carrots, the remembrance of what I called Charlock occurred to me, and I determined to experiment on the plant. With this view, I gathered in the fields, and as far as possible from allied plants, such as Cabbages, Turnips, Radishes, &c., seeds of Raphanus raphanistrum, and sowed them with the intention of resowing in several successive years, selecting every time seeds from those plants which presented the most promising features.

To give my experiment, which was continued during five successive years, a greater amount of certainty, and to impress upon it more deeply the seal of truth than would otherwise have been the case, it was carried out under two different conditions—viz., at Paris, in the light, dry soil of the nurseries of the Museum of Natural History, and in the country, in a strong calcareous clay. Under these two conditions the results obtained were nearly similar. At Paris the long form of root predominated, and was almost the only one, and in the country it was the reverse. Again, whilst at Paris, only white or rose-coloured roots were produced, in the country these were purple, or very dark brown verging on black, and there were some of all forms and colours.

The fact that different soil types yield different types of roots, though large and edible in both cases, strongly suggests that dual selection — for soil type and large roots — produced different results. Or, similar results but in different proportions. This is evidence that alleles favoring different soil-types are differentially linked to other genes (alleles) affecting the shape and color of the roots.

Then there is the work of Henry Eckford with Sweet Peas. He started with seven old and highly inbred strains, and in short order was introducing new colors and forms. One could insist that he was merely blessed with lots of random gene mutations that no one else had seen. Or, as I suspect, by selecting for larger flowers with broader petals, and more flowers to the stem, he brought out new expressions of genes that had long been masked.

Another interesting example can be found in Belyaev: Domestication of the Silver Fox (1979). Although he selected only for behavioral characteristics, and was careful to avoid inbreeding, his animals began to exhibit traits that had not been seen before in captive-bred foxes: long tail, short tail, curled tail, floppy ears, white piebald, brown piebald as well as disruptions in the breeding cycle.

Clausen &Hiesey (1960)
The arrangement whereby genes which govern a character are located on separate chromosomes provides a fair degree of resilience in the inheritance of the character, and in the adjustment of the race to minor differences in environment. From the opposite point of view, if the genic controls of characters were closely assembled on single chromosomes, recombination would involve large steps and the coherence-variation equilibrium would become impaired. It should be remembered, however, that genes located remote from each other on a chromosome can function as if they were independent of each other.