Ecological Genetics and Evolution: Essays in Honour of E. B. Ford (1971) pp. 21-26
Plant Evolution in Extreme Environments
A. D. Bradshaw

Artificial Selection Experiments

To understand how such evolution can come about we need to appreciate the power of selection. A starting point is the famous long term selection experiment, the Illinois corn experiment, which was started in 1900 and has continued ever since. Artificial selection for low and high oil and protein content has been carried out on an ordinary unselected open pollinated population, the Buff White variety of maize. The population size has been maintained at several thousands of plants and in each generation the twelve highest or lowest plants out of sixty analysed are selected to give rise to the next. The course of selection for oil content is shown in fig. 2.1. The significant feature is the long and continued response to selection which has taken the selected populations to levels of oil completely outside that of the range of the original population (Woodworth, Leng & Jugenheimer 1952). The rate of progress under selection has continued almost undiminished over the whole period.

FIGURE 21. The outcome of selection for high and low oil content in the Illinois corn experiment
(From Woodworth et al. 1952).

A similar experiment has been carried out in ryegrass Lolium perenne where flowering time has been selected (Cooper 1960) (fig. 2.2). There is the same picture of change although in this case the basic population is only 125 plants and only four are selected for the next generation. One aspect of the experiment is that in the first generation two different sets of parental plants were chosen as starting points. The effect of these has continued throughout.

FIGURE 2.2. The outcome of selection for heading date in ryegrass (from Cooper 1960).

The characters that have been selected in these two experiments are controlled by a large number of genes. The degree of response achieved concurs with what might be expected from genetic theory. It also agrees with work on Drosophila. Such patterns of change are likely with a series of additive genes, scattered through an initial population (Lee & Parsons 1968). In the restricted ryegrass populations the effect of choice of starting material is very evident and is similar to that found by Hosgood & Parsons (1967). There is little sign of restriction of release of variation due to linkage: this can be explained by chromosome numbers which are effectively at least twice that of Drosophila.

The Illinois corn experiment with its relatively large population is the model most relevant to what may happen in a natural population. Plant populations are usually large in size. And it must also be remembered that the size of a natural population cannot be measured solely in terms of the number of observed adjacent individuals: it must take into account individuals farther away supplying new genetic material by gene flow. In plants, gene flow due to pollen movement by wind or insects is leptokurtic, so that there is a small amount of gene flow over relatively large distances (Bateman 1947) sufficient to be important in supplying new variability (see p. 41).

Another way we can understand how a population changes under the influence of selection is to look at situations in which various sorts of artificial populations have been put into natural conditions, and the change of their constitution followed over a number of generations. One of the most interesting models was that set up by Charles (1961). In this three different genotypes of ryegrass L. perenne (in fact three different cultivated varieties) were mixed and sown in a field and subsequently subjected to a number of different treatments. After sowing the sward was sampled at regular intervals and the absolute and relative amounts of the three different genotypes determined. The rate of change (fig. 2.3a) is remarkable. In the course of a single year during which no sexual replacement has occurred frequencies of the components have altered by a factor of three, implying very high coefficients of selection. The degree of change depends upon the treatment imposed. The reason for the changes becomes clear when the absolute numbers of the components in the mixture are examined (fig. 2.3b). The original high numbers of seedlings fell markedly through the period of observation, allowing startling changes to occur in the relative frequencies of the components without any of them having increased in absolute number. This may seem to be a very artificial situation, but it is not. Plants are overgenerous in their seed production and the changes that occurred in this experiment are just the changes that could occur among the offspring of plants inhabiting a particular environment. In 1952 samples of a hybrid population of rice, formed from crossing two varieties Noren 20 and Zuiho which have very different dates of maturity, were set out in a range of environments extending from the extreme north to the extreme south of Japan (Akemine & Kikuchi, in Allard & Hansche 1964). Each year the hybrid populations were harvested and re-sown in their particular habitats. At the same time a certain amount of each material was taken to a central situation (Hiratsuka) and grown for observation. Figure 2.4 shows the changes in flowering time that were observed in comparison with the behaviour of the original parents used as controls. Over the course of four generations the populations in the extreme environments have come to resemble very closely the characteristics of the original parental varieties. The populations in intermediate habitats have taken on their own distinctive characters. Only one population, Hiratsuka, appears to remain in a very variable state after four generations.

FIGURE 2.3. Change in genotype frequency in a ryegrass mixture (from Charles 1961).   FIGURE 2,4. Histograms of heading dates in rice populations grown for successive generations in different parts of Japan (from Allard & Hansche 1964)

In 1937 a paper was published on the changes in the composition of the species of a pasture which had been sown three years previously in Maryland and divided into two halves, one of which was grazed and one of which was cut for hay (Kemp 1937). Very extensive changes in the characteristics of blue grass (Poa pratensis), orchard grass (Dactylis glomerata) and clover (Trifolium repens) were recorded. There was a much higher frequency of prostrate individuals of all three species in the grazed part than in the mown part. However, despite all the interest in the formation of ecological races or ecotypes in plant species, almost nobody appears to have paid any attention to the results or realized their significance. We have had to wait until Stebbins discussed them in 1950. It is perhaps very easy to dismiss the results as an accident or faulty observation. Yet taken now with the support of all the evidence for the power of selection that more recent experiments have given us, they appear very prophetic of the way in which natural populations can change. It is fascinating that the grass plots that have been differently fertilized in the Park Grass experiment at Rothamsted have recently been shown to contain radically different populations of Anthoxanthum odoratum differing not only in size and growth rate, but in growth habit, disease resistance and flowering time (Snaydon 1970).

At the same time as Kemp reported what happened in a pasture a paper appeared describing work carried out in Sweden on the effects of propagating varieties of herbage plants in regions away from where they were originally bred (Sylven 1937). One example was Bottnia meadow fescue which was bred for north Sweden. The effects on its fitness in north and south Sweden of propagating it for one generation in south Sweden is given in table 2.1; more radical changes in the characteristic of a population cannot easily be imagined. Now a great deal more evidence of this sort has accrued (e.g. Beard & Hollowell 1952) and plant breeders realize that shift in characters of varieties due to natural selection is a major problem in seed production.

TABLE 2.1. Changes in yielding ability of Bottnia meadow fescue (bred for grass production in north Sweden)
after one generation of seed production in north and south Sweden (Sylven 1937).

Present Site S. Sweden N. Sweden
Seed previously grown in N. Sweden S. Sweden N. Sweden S. Sweden
Yield (Tonnes/Ha) 44.4 109.1 58.9 55.9
per cent change from original 100 245.7 100 94.9

When Fisher wrote The Genetical Theory of Natural Selection in 1930, he envisaged selective advantages of up to about 1 per cent. This attitude to the power of selection prevailed for a long time and is probably why Kemp's and Sylven's results were disregarded. Another reason is that the study of plant evolution has been dominated by a taxonomic approach: the main concern has been to document distinctive populations, or ecotypes. But now attitudes to selection have changed, and Ford (1964) in particular has given us overwhelming examples of much higher selective advantages in animals, so that we realize that we must look at animal populations with a very different perspective. In plants too we now have plenty of examples showing how far Ford's arguments must be applied to plant populations. The phase when we can think of plant populations as more or less static entities is over. We are in a good position to find out how evolution proceeds in natural plant populations in natural situations.