Heredity (1973) 30 (1), 33-40
SELECTION FOR SPECIFIC AND GENERAL RESPONSE TO ENVIRONMENTAL DIFFERENCES
J. L. JINKS and V. CONNOLLY*
* Present address: Plant Breeding Department, The Agricultural Institute, Oak Park, Carlow, Ireland.
Department of Genetics, University of Birmingham, Birmingham BIS 2TT, England
The extent to which it is possible to predict the performance of selections, made on the basis of their performance in one environment, when grown in other environments and the modifications of the designs of selection programmes that are necessary in order to select for either indifference or sensitivity to environmental variation as well as for mean performance, have been investigated using rate of growth of the dikaryotic stage of the heterothallic, basidiomycete Schizophyllum commune as a model system. The material was the eighth generation of 10 selections. Eight of these, two selected for high and two for low rate of growth from each of two wild isolates, consisted of one high and one low selection from each isolate selected on their performance at 20°C. and the other high and low selection from each isolate on their performance at 30°C. The remaining two selections, one high and one low selection from one of the two isolates, were selected on the basis of their performance at both 20°C. and 30°C. These 10 selections and the two original, unselected wild isolates were compared for their rates of growth over nine temperature environments covering the range 15°C. to 35°C. at 2.5°C. intervals. Lines selected from the same isolate in the same direction but on the basis of their performance at different temperatures, showed significant genotype-environmental interactions with the nine temperature environments. These interactions, which were analysed using both dependent and independent assessments of the environments, could be directly related to the temperatures at which they were selected. In general, selections made at the lower temperature deviated more in the direction of selection at lower temperatures than selections made at the higher temperature, while selections made at the higher temperature deviated more in the direction of selection at higher temperatures than selections made at the lower temperature. Selections made on the basis of their performance in two different temperatures maintained their deviations in the direction of selection over extreme temperature environments much better than selections made at a single temperature. Selections for a low rate of growth at a temperature that leads to a relatively high rate of growth gave lines that were less sensitive to temperature changes than for the same selection made at a temperature that leads to a relatively low rate of growth. The reverse was true of selections for high rate of growth. Selections made on the basis of performance at two different temperatures were always intermediate in their sensitivity to temperature changes. Since all these results are predictable from a simple and general model of gene and environmental action and interactions, they presumably have a general validity.
WITH the growing awareness of the importance of genotype-environmental interactions for continuously varying characters, it is becoming increasingly important to know more about the consequences of such interactions on the progress and end-products of selection programmes. For example, to what extent can we predict the consequences of selecting in one environment on the performance of the selections in that environment relative to the performance in other environments, and what modifications must we make to the design of selection experiments in order to select for indifference or sensitivity to environmental variation as well as for mean performance? The existing evidence was recently summarised by Bateman (1971), who claims that the balance of evidence suggests that more widely adapted genotypes will generally be produced by selection in the less favourable environments. In this paper we shall present the results of selection experiments on the fungus Schizophyllum commune. On the basis of these results and simple theoretical models we put forward a general interpretation that also accounts for the similar results of comparable selection experiments with mice (Falconer and Latyszewski, 1952).
Schizophyllum commune, a heterothallic basidiomycete, has proved to be a technically ideal organism for studying continuously varying characters (Simchen and Jinks, 1964; Simchen, l966a; Simchen, 1967), the response to selection for such characters in a constant environment (Simchen, 1966b, Connolly and Simchen, 1968) and the interaction of the genotypes with deliberately imposed environmental treatments (Fripp and Caten, 1971; Fripp, 1972). It is, therefore, an obvious choice for preliminary and rapid investigations of the effect of genotype-environmental interactions on the progress of selection.
2. MATERIALS AND METHODS
The material is the eighth generation of 10 selection lines and the original unselected dikaryotic isolates, 2 and 6, from which the selections were derived. The origin and properties of isolates 2 and 6 have been described by Simchen and Jinks (1964) and Simchen (1966) and the properties of some of the selection lines and their progenies have been described by Fripp and Caten (1971) and Fripp (1972).
The selections were for high or low rate of growth, measured as the total linear growth in millimetres over 9 days on 2 per cent. malt agar (see Simchen and Jinks, 1964, for details). For each of the two isolates a high and low selection was made to 20°C. and at 30°C. Two further selections were made on isolate 2 only and these were for high and low rate of growth, the selection in each generation being on the basis of average growth at both 20°C. and 30°C. Apart from the temperature differences all other conditions, e.g. replication, family sizes and selection pressures, were constant: for all selection lines in all generations. In each generation of selection 50 dikaryons of each selection line were assessed for their rates of growth by growing two replications of each dikaryon, one in each of two independently randomised blocks. These 50 dikaryons were obtained by mating randomly chosen pairs of monokaryotic progeny of the two dikaryons selected in the previous generation, one of each mated pair coming from each of the two selected dikaryons. The most extreme pair of dikaryons in the direction of selection from among the 50 assessed for their rates of growth were then chosen as the parents of the next cycle of selection. Thus 4 per cent, were selected at each generation in each line.
For the high and low selections of isolate 2 which were based on the rate of growth at both 20°C. and 30°C. the data were rescaled before choosing the extreme pairs of dikaryons to be parents of each of the two selection lines. This was necessary because the growth at 30°C. was about double than that at 20°C. The rescaling consisted of dividing the rate of growth 0f each dikaryon in each temperature by the average rate of growth of all dikaryons at that temperature. In this way growth at each temperature on average contributed equally to the final index upon which the selection was based, thus avoiding the situation where a dikaryon with a very high rate of growth at 30°C. but only an average rate of growth at 20°C. might be selected because it gave the highest overall rate of growth on a straight average (Connolly and Simchen, in press).
The eighth generation of the 10 selections and the two original isolates were compared for their rates of growth on 2 per cent. malt in nine temperatures over the range 15°C. to 35°C. at intervals of 2.5°C. Duplicates of each of the 12 lines were grown as single randomised individuals in each of two independently randomised blocks in each temperature. For the selections the duplicates were the two most extreme dikaryons chosen from the seventh generation of selection. For the original isolates the duplicates were two independent clones of each of the two isolates.
An analysis of variance, of the growth rates of the 12 lines grown in the nine environments is given in table 1. This shows highly significant differences between the lines and between the environments as well as significant interactions of lines with environments. There are no differences between duplicates but there is a significant interaction of duplicates with environments. The latter, however, is relatively small and the interaction of lines with environments is significant when tested against it. The significant duplicate x environment interaction can be traced to differences between duplicate low selections and in particular to the low selections of isolate 2.
Analysis of variance of rate of growth of the 12 lines grown
in nine temperature environments
|Blocks within E||9||6.95||n.s.|
|Duplicates within L (D)||12||39.69||n.s.|
|L x E||88||425.47||<0.001|
|E x D||96||29.64||<0.001|
|Blocks x L||99||9.99||n.s.|
|Blocks x Duplicates||108||9.69||--|
|Pooled Block interactions||207||9.83||--|
|*Significance when tested against pooled block interactions except for D and L x E which are tested against E x D.|
The line differences and their interactions can be partitioned in a number of ways, but the most informative for our present purposes is to compare selections from the same isolate that have been selected in the same direction but in different environments. That is, to compare the lines within the following four groups:
|the three low selections of isolate 2;|
|the three high selections of isolate 2;|
|the two low selections of isolate 6; and|
|the two high selections of isolate 6.|
The four analyses of variance of lines and environments show highly significant lines x environment interactions within each of these four groups (P <0.01 in each case). Thus lines which have been identically selected from the same starting material but in different environments show significantly different changes in performance with changes in the environment. We shall now show that these differences can be simply and directly related to the environments in which the selections were made.
The mean rates of growth of each of the 10 selections and of the two original, unselected isolates in each of the nine environments after averaging over duplicate lines and replicates are given in table 2. If we confine our attention for the present to selections made either at 200 C. or at 30°C., the results in table 2 show that at 20°C. the lower of the two low selections of both isolates 2 and 6 are those selected at 20°C. Similarly, the higher of the two high selections of both isolates at 20°C. are those selected at 20°C. At 30°C. the situation is reversed and the lower and higher of the low and high selections for both isolates are those selected at 30°C. These relationships, however, are not confined to the temperature at which the selections were made but extend to neighbouring temperatures. For example, the low selection of isolate 2 selected at 20°C. is lower than that selected at 30°C. over the range 15°C. to 27.5°C. while the reverse relation. ship holds from 30°C. to 35°C. Similarly, the high selection of this isolate, selected at 20°C., is higher than that selected at 30°C. over the range 15°C. to 25°C. but this relationship is again reversed from 27.5°C. to 35°C. A less consistent, but similar, set of relationships holds for the corresponding selections of isolate 6. A more extreme temperature dependence is shown by some of the high selections which are slower growing than the original, unselected isolates in the extreme temperatures. This is most marked for high selections made at 30°C. when grown at the lowest temperature and for the high selections made at 20°C. when grown at the highest temperature (table 2).
If we now extend the comparisons to include the selections of isolate 2 which were based on the average rate of growth at 20°C. and at 30°C. we find that the low selection is in general intermediate between the low selection made at 20°C. and that made at 30°C., but it is the lowest of the three at the two extreme high temperatures. The high selection based on rate of growth at both temperatures, on the other hand, is clearly the highest of the three high selections of isolate 2 at all temperatures except 20°C. and 30°C., when its performance is close to the selections made solely at one or the other of these temperatures. The selections based on the performances in both temperatures are, therefore, as good as, or better than, the average of the corresponding selections made in one or the other of the temperatures and they maintain their performance over a wider range of temperatures.
Mean rate of growth of the 12 lines in the nine temperature environments expressed as linear growth in millimetres
over 9 days after averaging over duplicate lines and replicate blocks
|Isolate 2||Isolate 6|
|Unselected||Low selections||High selections||Unselected||Low selections||High selections|
An alternative way of looking at these data is through the regression analyses which have been reviewed by Perkins and Jinks (1968), Freeman and Perkins (1970) and Fripp (1972). There are a number of ways in which the regression analyses can be carried out on these data and the results of two, which appear to be particularly informative, will be presented here. The first method is a joint linear regression analysis among the lines within each of the four groups described earlier (I to IV, page 36) using the mean over lines within each group in each environment to provide an estimate of the environmental value, ej, (j=1 to 9) for each group. In the second method the 10 selection lines have been divided into two groups according to whether they originated from isolate 2 or isolate 6 and the mean of isolate 2 or 6 in each environment provides an independent estimate of the environmental value, zj, for the two groups of selections, respectively.
Apart from the lines of group IV (the two high selections of isolate 6) the linear regression of performance of each line in each environment on the corresponding environmental value (ej or zj) account for a proportion of the genotype-environmental interactions that is significantly greater than the non-linear residual variation. The regression coefficients () obtained for each selection line by the two methods of analysis are summarised in table 3. The values of obtained for each line by the first method are relative to those of the other lines in the same group. The corresponding values from the second method are relative to the unselected original isolates 2 and 6. While, however, this leads to large differences in the absolute values of 's between the two methods, the relative values within the four groups are quite consistent.
|Linear regression coefficients ('s) for the regression of the genotype-environmental interaction components of rate of growth of the selection lines grown in nine environments on dependent estimates (method 1) and independent estimates (method 2) of the environmental values. There are significant differences between these values for the lines within each group except for the two lines of Group IV
|Selection line||Method 1||Method 2|
|Group 1||Low 20°C.||0.3l||-0.38|
|Isolate 2||Low 30°C.||-0.25||-0.7l|
|Low 20°C. and 30°C.||-0.06||-0.59|
|Group II||High 20°C.||-0.09||0.01|
|Isolate 2||High 30°C.||0.07||0.19|
|High 20°C. and 30°C.||0.02||0.l5|
|Group III||Low 20°C.||0.34||-0.53|
|Isolate 6||Low 30°C.||-0.34||-0.78|
|Group IV||High 20°C.||0.06||0.3l|
|Isolate 6||High 30°C.||-0.06||0.28|
An obvious conclusion that can be drawn from table 3 is that where selections made at 20°C. and at 30°C. differ in their linear sensitivity to the environmental differences, the low selection made at 30°C. and the high selection made at 20°C., are the less sensitive of the two low and two high selections of each isolate, respectively. Equally, the high and low selections of isolate 2 selected on the mean performance at 20°C. and at 30°C. are intermediate among the high and low selections of this isolate in their linear sensitivity to the environmental differences. These rankings of the selection lines remain unchanged if we consider their total variation over environments instead of their linear change with environment (cf. Falconer and Latyszewski, 1952).
In the previous section we have established that lines which have been identically selected, except that the selections were made at different temperatures, respond in different ways to changes in temperature. We have identified consistent relationships between the selections made at different temperatures that make sense only if we assume that in each temperature used for selection we have fixed some alleles which are fully expressed only at these or similar temperatures. Thus all the low selections are equally likely to have fixed those alleles which give below-average rate of growth at all temperatures. The low selections made at 20°C. will also have fixed in addition those alleles that have an average or aboveaverage rate of growth over all temperatures combined with an aboveaverage sensitivity to temperature changes so that they confer below-average rate of growth at 20°C. but above-average at higher temperatures. In contrast, the low selection made at 30°C. will have fixed instead those alleles that have an average or above-average rate of growth over all temperatures combined with a less than average sensitivity to temperature changes so that they confer less than average rate of growth at 30°C. but above-average at lower temperatures. Hence, we expect and observe that the low selection made at 30°C. is the less sensitive to temperature changes. For the high selections the arguments are reversed and we both expect and observe that the high selection made at 20°C. is the less sensitive to temperature changes.
The low and high selections made on the basis of the mean rate of growth at both 20°C. and 30°C., in addition to fixing alleles that give below and above-average performances over all temperatures are most likely to fix alleles which give consistently below- or above-average performances at both temperatures, respectively. Consistency of performance means average sensitivity to temperature changes. It fits the expectations, therefore, that these selections of isolate 2 are intermediate among the three low and three high selections in their sensitivity to temperature changes.
In that the results are consistent with those expected on a simple and general model of genotypic and environmental action and interaction we can draw a number of conclusions that should have general validity.
1. There are relatively small differences in mean performance when taken over the whole range of temperature environments among lines selected in the same direction but in different temperature environments. There are, however, relatively larger differences among such lines in their performance in specific temperature environments that are related to the temperature at which they were selected; selections made at the lower temperature deviating more in the direction of selection at lower temperatures than selections made at higher temperatures, and vice versa. Hence, to achieve a desirable average performance over a wide range of environments, selection in any one of a number of environments that are well within this range appears to be equally satisfactory. To achieve an equally desirable performance in one specific environment, selection must be practised in that or in a closely related environment.
2. Selections made on the basis of mean performance in two different temperature environments maintain their deviations in the direction selected over extreme temperature environments much better than selections made at a single temperature. Hence to achieve a desirable average performance over a range of environments including an acceptable performance even in the worst of these environments, selection must be based on average performance in two or more contrasting environments within this range.
3. When the direction of selection and the effect of the environment on selection are opposed in their effects on the deviation of the resulting phenotype from the mean phenotype, e.g. selection for a low rate of growth at a temperature that leads to a relatively high rate of growth, the resulting selection lines are less sensitive to environmental variation than when direction of selection and the effect of the environment of selection are reinforcing, e.g. selection for a low rate of growth at a temperature that leads to a relatively low rate of growth. Hence, selection for mean performance will lead to greater or lesser sensitivity to environmental change within a range of environments according to whether the environment in which selection is carried out deviates from the mean of the range in the same or in the opposite direction to the direction of selection.
Acknowledgments.—We are indebted to Dr G. Simchen for his interest throughout the course of this work and to Professor K. Mather for his helpful comments during the preparation of this paper.
CybeRose Note: The information presented in this article is extremely valuable, though presented in a way that makes it almost unitelligible to anyone who is not a statistician. I will try to translate the conclusions into English.
1. Strains selected for uniform growth at low temperatures are more uniform in their growth at low temperatures than at high. Strains selected at high temperatures are more uniform at high temperatures than at low. Average performance over the full range of temperatures is roughly the same for all strains selected at a single temperature.
2. Strains selected at both high and low temperatures are more uniform in their growth over the full temperature range than any strain selected at a single temperature. Two selection temperatures are better than one.
3. The fungus studied ordinarily grows faster at high temperatures than at low. Selection for slower growth at high temperatures reduced variance more than selection for faster growth at high temperatures. Likewise, selection for faster growth at low temperatures led to more uniform performance than selection for slower growth. Going against the grain, so to speak, gives more uniformity than going with the grain.
Below is a graph of the data from Table 2 for low selection and high selection of isolate 2. It shows that selection for high growth rate at one temperature also raises the average growth rate over the full temperature range, though most notably in the high temperature range.
I think it would be very interesting and informative to repeat the experiment with an additional selection method: faster growth at 20°C and slower growth at 30°C. The fungus obviously possesses the "genes" necessary for both responses, so it would only require organizing them into a stable growth rate over a broader range of temperatures.
The practical implications are endless. It is worth noting that this is precisely what the Michurinists were discussing decades earlier when they were attacked as "unscientific" by the neo-Darwinists (such as K. Mather). Expression of a trait can vary with the environment. If we want to stabilize a trait over a wide range of environental conditions we must make our selections in at least two distinct environments.
The experiment could be varied with other environmental conditions. For example, a gradient of soil type could be prepared, ranging from pure sand to rich garden loam. In this way a plant breeder could develop plants that will perform acceptably in different conditions.
Or, we could test for responses to varying concentrations of salt, copper, petroleum waste, etc. to identify selections that are more tolerant, and observe the improvements brought by selective breeding.