When we begin to study heredity we are treated to the familiar "Pea Laws" of Mendel. Rough vs. smooth, green vs. yellow; well-behaved dominant and recessive traits segregating freely, uncontaminated by other genes. Plant breeding is reduced to Punnett squares and probabilities. Or so we are assured by textbook writers.
When we take brush in hand and begin pollinating on our own we quickly learn that heredity is not such a simple matter. Red x white may give numerous shades of pink, as well as yellow and purple and orange in seemingly unpredictable assortments. Le Grice, for example, crossed the pink-flowered R. californica with white garden roses and found browns and purples in the progeny. Clearly simple Mendelism doesn't have all the answers.
A single trait may involve multiple genes. And a single gene will have multiple effects -- some dominant, some recessive. It is incorrect to speak of dominant and recessive genes when the same gene has both dominant and recessive effects at the same time.
Most traits are influenced by a multitude of genes. Some of the effects are strong, some minor, and yet others are individually insignificant but may override the effects of a major gene when several (or many) work together.
The Russian geneticist, Philiptschenko, spent several years analyzing wheat to identify 6 pairs of genes that combine to produce the distinctive broad grains and glumes of the Marquis variety. The major gene proved to be recessive in its effect, two secondary genes had dominant effects, and the remaining 3 were also recessive. Thus, the true-breeding Marquis carries the genes [aa+BB+CC+dd+ee+ff].
If a strain of Marquis wheat is crossed with a another strain [AA+bb+cc+DD+EE+FF], the F1 will have the formula [Aa+Bb+Cc+Dd+Ed+Ff], and will have ordinary, narrow grains.
About one-fourth of the F2 seedlings will carry the major gene pair aa, but the expression of the Marquis trait will be variously diminished. This is called the "Engeldow Shift". Mendelism promises that recessive traits will be recovered in the second generation, but what we get back may not be what we put in. The 'a' gene has not been altered, but its expression is different in different genetic environments. In other cases the expression of a trait might be enhanced.
Philiptschenko's model also allows us to examine modification of dominance.
From the hybrid we may select two strains of wheat that differ from the Marquis in one gene pair each. In the first strain aa is replaced with AA, giving (AA+BB+CC+dd+ee+ff). When this strain is crossed with the original Marquis, all the offspring will be homozygous for (BB+CC+dd+ee+ff) and heterozygous for Aa. So the expression of the trait will be controlled by A or a. That is, the Marquis trait will prove to be recessive to narrow-grained, and will be fully expressed in 25% of the F2.
For the second strain, bb replaces BB (aa+bb+CC+dd+ee+ff). Crossed with the Marquis, this gives F1 hybrids heterozygous for Bb, but otherwise homozygous. Thus, the Marquis trait will behave as a Mendelian dominant, giving 100% Marquis in the F1, 75% in the F2.
These two examples show that a polygenic trait may be either dominant or recessive even though the same genes are involved, and both strains are derived from the same original population.
In simple Mendelian inheritance there is a neat segregation of unit traits in the F2 generation. In some plants red x white may give all pink in the F1, but in next generation there will be red and pink and white in predictable proportions -- 1 red : 2 pink : 1 white. According to the textbooks.
In practice the precise proportions are rarely obtained. Larger samples allegedly give closer approximations, but even this assumption is incorrect. In fact, as much research has demonstrated, different alleles typically have effects other than those being studied; and these other effects may influence the viability and vigor of pollen and ova, of embryos and of the developing organisms.
Albinism in animals, for example, affects more than pigmentation. In many cases vision is impaired and growth rate is reduced relative to the normally pigmented siblings. When an albino Betta (Siamese Fighting Fish) was mated with a normal female, and the offspring were mated, Mendelian calculations would lead us to expect that about 1 of 4 would be albino. In fact there were none.
A larger progeny would not improve the survival rate of the albinos. Bettas spawn for 4 to 5 hours, and the eggs hatch over a similar period. The first fry to hatch begin feeding sooner, and grow so fast that they can eat their younger siblings.
The albinos develop more slowly, and are thus smaller than siblings of the same age. When they hatch, their poor eyesight limits their ability to feed and grow. It is no surprise that the albinos were selectively eliminated from the progeny. Fish food!
In plants we meet similar variations in survival rate. Pollen tubes may grow at different rates depending upon the genes they carry. Self-sterility genes may mark certain pollen for rejection by the stigma or destruction by the style. Embryos with different gene combinations may develop at different rates and vary in their ability to demand nutrition from the seed parent. In some cases there is selective elimination or promotion of certain genotypes in the ovules.
These differences may not all be due to the traits being observed. Genes that are coincidentally linked can also influence the rate of survival. So, if we self-pollinate the various F1 offspring, and raise their progeny separately we may find differences in the proportions of the various genotypes.
Traditional Mendelists would add together all the F2 progeny, looking for confirmation of the predicted ratios. But if we resist the temptation to be traditional, we may find that the deviations from theory are consistent. For example, one of the F1 plants may produce a noticeable deficiency of whites while the pinks and reds are proportionately "normal". That is, for example, 0.5 white : 2 pink : 1 red for an acceptably large progeny. A different plant may give a disproportionately large number of pinks (heterozygotes); e.g., 0.7 white : 2 pink : 0.5 red. Finally, there may be the rare plant that is deficient in pink offspring.
If we start with one of the populations that gave deficiencies of both types of homozygotes, we could strengthen the deviation by continuing to select for the pink flowered plants which give the fewest whites and reds. After a few generations the strain would probably become stable, but continued selection over many generations could improve the strain even further.
Harmful recessive mutations are common -- more common than beneficial mutations, and far more common than dominants. Since, in this example, we are favoring the heterozygotes (pinks) we are also protecting any harmful mutations that are recessive in their effects and linked to the genes that are always heterozygous. And since we are also favoring pinks that give the highest proportion of pinks, we are indirectly favoring harmful recessive mutations.
Eileen Erlanson raised self-pollinated seedlings of wild roses and found that some of the seedlings were remarkably prone to mildew or rust. We might have assumed that disease resistance would be such a desirable trait that resistance would long ago have become fixed in the races. That this is not the case suggests that the alternative to rust resistance, for example, must also be of use.
It may be that the alternative allele gives resistance to a different disease or a different strain of the same disease. This has been found in humans and other organisms.
Pathogens have their own distributions, climate preferences and cycles of activity. For several years environmental conditions may favor one pathogen, and resistant roses could survive. Those homozygotes with double resistance could out-reproduce both the heterozygotes and those homozygous for the alternative resistance.
If the conditions continued for decades, the relative frequency of the two types of resistance would be shifted to favor resistance to the currently active pathogen.
Suppose that the alternatives are resistance to mildew (M) and resistance to rust (R). The possible genotypes in this hypothetical case are MM, MR and RR. Without genetic restriction on homozygotes, the mildew years would favor a shift in the population towards MM. MR types would survive, but the RR would be heavily reduced.
In time, though, the climate would shift again. The strongly M-biased population could be devastated by a rust epidemic. The population could then shift back towards MR and RR, but not without heavy losses that could also affect other species. Bees feed on pollen, birds on hips. Catastrophic changes in the rose populations could affect the availability of birds and bees, which would then restrict the roses' ability to refill their allotted territory. Fewer roses will attract fewer bees, and fewer hips will attract fewer birds.
Thus it becomes highly desirable to restrict the production of homozygotes. MR types might not be so resistant to mildew as MM, and not so resistant to rust as RR. But the MR type could survive both diseases, and any MR types that produce fewer MM and RR offspring will have a long-term advantage in a region of fluctuating climate. (Modification of Dominance could make MR as resistant to rust as RR, and as resistant to mildew as MM. But that's another story.)
When we see a stunted and diseased specimen, we naturally assume that the disease is responsible for the stunting. This may not be the case. The stunting may be due to linked semi-lethal genes that normally prevent the homozygotes becoming numerous and threatening the long-term survival of the population.
Disease prone plants of normal vigor may represent "delinkages". That is, a crossover may have occurred within the linkage group separating the resistance gene from the normally associated set of recessive harmful genes. R, for example, could become linked to the restricting genes usually associated with M. When this R is paired with a "normal" R (with the usual set of harmful or restricting genes), the plant is of normal vigor, very prone to mildew but possibly very resistant to rust.
In this hypothetical example I have assumed that the mildew resistance is the alternative to rust resistance, which may not be the case. It might be demonstrated by crossing mildew-prone and rust-prone siblings from self-pollinated wild plants. It is also possible that susceptibility to disease has been favored to protect some other heterozygous pairing.
It is frequently claimed that inbreeding leads invariably to homozygosity. This is not always the case.
Harmful recessive mutations are common, and are ordinarily eliminated from a population when homozygous. But where heterozygotes are selectively favored, any recessive lethal or semi-lethal mutations which happen to be linked to the heterozygous genes will also be protected from elimination. These harmful recessives further reduce the occurrence of homozygotes. Thus, once such a system of protected heterozygotes has been established, it tends to be self-supporting.
External selection may favor the original heterozygous pairing, but accumulating recessive mutations can turn the selection inwards by reducing the viability of the homozygotes.
As I explained above, external selection for heterozygotes can lead to internal elimination of homozygotes. There is no mystery or magic involved. Dust bunnies are more common under beds because that is where they are most protected.
An important consequence of selection for heterozygotes is that the advantage tends to increase over time. Whether selection is Natural or Artificial, external elimination of homozygotes (even slight) favors accumulation of genes that are harmful when homozygous but whose effects are recessive in the favored heterozygotes.
As the homozygotes become increasingly crippled by the accumulating harmful genes, the heterozygous advantage is relatively increased.
The "harmful" genes are not necessarily without advantages of their own. Any gene may have a variety of effects, some masked by other genes (recessive) and others active even when heterozygous (dominant). It is not just a matter of what genes are present in the opposing members of a heterozygous pairing, but how they are arranged.
Two non-homologous genes may counter each other's harmful effects. If they happen to be linked on the same chromosome, they could be homozygous or heterozygous with their alternatives and in no case reveal their potential for harm. But when separated onto different homologous chromosomes their harmful effects will be neutralized only when both genes are present in the same organism -- in heterozygotes.
That is, a and b may be harmful only when one is homozygous and the other is absent. AAbb and aaBB might be lethal combinations while AaBb survives due to the balancing of a against b. If a and b are linked, then ab and AB will segregate as units and the harmful effects will appear only in rare "mutations" when a crossover separates the linked genes.
Now, if selection happens to favor a heterozygous pairing of genes located near our ab and AB pairs, crossovers may separate a from b without harm.
Suppose that pink is favored over red or white. Further suppose that ab is coincidentally linked to red while AB is linked to white: Rab and WAB.
If a crossover separated a from b we could have RaB and WAb. Since selection is already favoring RW, aB and Ab will be protected from elimination, and will further reduce the frequency of homozygotes.
In the above example we assumed that ab were already linked to each other and to the "gene for red". In the Real World the interactions among genes are far more complex than in this simple model. One presents simple models as demonstrations at the risk of seeming simplistic.
One thing that must be mentioned is that crossovers are not purely random in their occurrence. Certain genes are known to inhibit crossovers in their vicinities. Other nucleotide patterns presumably also reduce or increase the frequency of crossover at specific locations. This matter is of considerable interest and importance, but there has been surprisingly little research into how it works.
Detlefsen and Roberts bred fruitflies with different pairs of traits; white eyes and miniature wings. Originally the crossover frequency was 36%, but through selective elimination they reduced the frequency to 6% in one strain, 0.6% in another. Crossover frequency is clearly under genetic control, and is thus subject to selection.
In our hypothetical example a and b were originally linked. So long as they were useful together, and not regrouped with A and B, natural selection would have acted to restrict crossovers between them without regards to W or R.
When selection came to favor WR over the homozygotes, selection against aB and Ab was removed. And because these new combinations supported the existing system of heterozygosity the new linkage groups RaB and WAb would be favored. That is, crossovers would come to be inhibited over the whole region containing the 3 gene pairs, rather than just over the shorter region of the 2 pairs.
No new mutations are required, except in the sense that aB and Ab would ordinarily be viewed as mutations. A change in the "selective pressure" can turn the rare into the common and redistribute localized crossover inhibition to form new linkage groups or supergenes.
If new mutations occur within the protected regions they may not only survive, but increase the effectiveness of the crossover inhibition by making the homozygotes even less able to survive.
Occasionally a segment of DNA will become inverted within a chromosome. When inverted and non-inverted regions are paired, crossovers are virtually impossible within the region. If they occur the resulting gametes are usually eliminated. So, in the overwhelming majority of cases the inverted and non-inverted segments behave as hereditary units -- supergenes.
Such inversions are surprisingly common and may soon be eliminated because the elimination of crossovers also results in a reduction in fertility. But when an inversion happens to coincide with a developing supergene, or some useful group of genes, there is an immediate leap forward in isolating the contained genes from their counterparts on the uninverted partner.
A supergene may be protected by localized crossover inhibition, by accumulation of harmful recessives, or by an inversion. Any or all may be active at the same time. Small inversions may be contained within a segment that also carries crossover inhibitors and recessive lethals. In fact, an inversion may also contain harmful recessives which thus prevents it from becoming homozygous. Likewise for the uninverted partner. And within the inversion there may be localized crossover inhibition, a relict of pre-inversion selection.