| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Dep. of Crop Sciences, Univ. of Illinois, Urbana, IL 61801. This research was supported by the Illinois Agric. Exp. Stn. and a grant from Renessen, LLC
* Corresponding author (jdudley{at}uiuc.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONResultsExplanation of Long-term...ConclusionsREFERENCES |
|---|
Abbreviations: IHO, Illinois High Oil • IHP, Illinois High Protein • ILO, Illinois Low Oil • ILP, Illinois Low Protein • QTL, quantitative trait locus (loci) • RHO, Reverse High Oil • RHP, Reverse High Protein • RLO, Reverse Low Oil • RLP, Reverse Low Protein • SHO, Switchback High Oil • SIM, simple interval mapping • SMA, single marker analysis • SNP, single nucleotide polymorphism
|
ACKNOWLEDGMENTS |
|---|
Received for publication July 9, 2007.
Dep. of Crop Sciences, Univ. of Illinois, Urbana, IL 61801. This research was supported by the Illinois Agric. Exp. Stn. and a grant from Renessen, LLC
* Corresponding author (jdudley{at}uiuc.edu).
Divergent selection for oil and protein concentration in the corn (Zea mays L.) kernel was initiated at the University of Illinois in 1896 by C.G. Hopkins. In 2005, 106 generations of selection had been completed for high oil and 105 for high protein. Limits to selection for low oil and low protein were reached but not for high oil or high protein. Over the more than 100 yr of the existence of the program a number of attempts have been made to analyze the experiment using quantitative genetic tools. The purpose of this paper is to trace the use of quantitative genetic techniques to analyze the results of divergent long-term selection for oil and protein, to relate results to the question of the need for divergent parents for quantitative trait locus (QTL) analysis, and to provide a new look at the reasons for long-continued progress from selection. Key findings include (i) progress from selection was much greater than could have been predicted; (ii) based on both classical quantitative genetic analysis and QTL studies, a large number of QTL are involved in control of the three traits; (iii) the number of QTL identified in a given study cannot be predicted by the magnitude of genetic variance or the divergence of the parents but is a function of the number of markers used and the number of lines evaluated; and (iv) epistasis may be an important factor in explaining long-term response to selection.
Abbreviations: IHO, Illinois High Oil • IHP, Illinois High Protein • ILO, Illinois Low Oil • ILP, Illinois Low Protein • QTL, quantitative trait locus (loci) • RHO, Reverse High Oil • RHP, Reverse High Protein • RLO, Reverse Low Oil • RLP, Reverse Low Protein • SHO, Switchback High Oil • SIM, simple interval mapping • SMA, single marker analysis • SNP, single nucleotide polymorphism
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONResultsExplanation of Long-term...ConclusionsREFERENCES |
|---|
The tools of quantitative genetics include means, regression analysis, variance estimation (particularly genetic variances), genetic correlations, mating designs, random mating, and more recently QTL analysis. The purpose of this paper is to trace the use of quantitative genetic techniques to analyze the results of divergent long-term selection for oil and protein concentration in the corn (Zea mays L.) kernel at the University of Illinois, to relate results to the question of the need to use divergent parents for QTL analysis, and to provide a new look at the reasons for long-continued progress from selection.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONResultsExplanation of Long-term...ConclusionsREFERENCES |
|---|
Classical Quantitative Genetics
Means
For the duration of the experiment, means of ears sampled each cycle are plotted against generations in Fig. 1
and 2
for the oil and protein strains, respectively. The earliest report of results from this experiment (Hopkins, 1899) compared means of IHO cycle 2 with ILO cycle 2 for oil and IHP cycle 2 with ILP cycle 2 for protein. For the IHO-ILO comparison, a difference of 11.5 g kg–1 oil was found while for the IHP-ILP comparison, the difference in protein was 5.0 g kg–1. From these results Hopkins concluded that selection for oil and protein was effective even though there was no test of significance of the observed differences. After 10 generations of selection Smith (1908) found a difference of 56.2 g kg–1 protein between IHP and ILP and a difference of 47.1 g kg–1 oil between IHO and ILO. These results were based on means of the ears grown in 1906 in each strain and again there was no statistical test of the significance of differences between means. After 28 generations of selection the difference in means between IHO and ILO was 83.5 g kg–1 oil and the difference between IHP and ILP was 82.2 g kg–1 protein (Winter, 1929). After 50 generations of selection the difference between IHO and ILO was 143.5 g kg–1 oil and the difference between IHP and ILP was 145.4 g kg–1 protein (Woodworth et al., 1952). Results of two cycles of reverse selection (selection within a strain in the opposite direction, e.g., selection for low protein in IHP) showed apparent progress in RHO, RHP, and RLP but little progress in RLO.
|
|
The first comparison of different cycles of the selection experiment grown in the same environments was reported by Dudley et al. (1974). Generations 65 to 70 of the forward selection strains were compared as were generations 19 to 22 of the reverse selection strains. Earlier generations were not included because seed was not available. Progress was significant and in the appropriate direction for all strains.
At generation 90, progress measured in replicated trials comparing the last 20 to 25 generations of selection of the forward and reverse strains was similar to that measured in each individual cycle (Dudley and Lambert 1992). No progress in either ILO or ILP was observed after generation 76. After 100 generations, results of replicated trials comparing the last 33 generations of selection showed continued progress for oil in IHO, RHO, RLO, and SHO (Dudley and Lambert 2004) and for protein in IHP, RHP, and RLP.
Means are an important measure of progress and an important statistical tool in all the work which has been done to help understand the genetic control of oil and protein. Mean values for individual ears were the units of selection. Means have been used in conjunction with other quantitative genetic tools in work following that of Smith (1908). Thus, further discussion of means will be included with discussion of variances or other quantitative genetic tools.
Variances
Winter (1929) was the first to report the effect of divergent selection on standard errors and coefficients of variation (CVs). He showed that selection in the high direction for either oil or protein decreased the CV among ears within strains while selection in the low direction increased it. Conversely, selection in the high direction increased the among ears within strains standard deviation while selection in the low direction decreased it. Based on these results, Winter (1929) concluded that continued progress in the high direction should be possible for both oil and protein but that a selection limit was being approached in the low strains. Woodworth et al. (1952) noted the same trends in standard errors and CVs within strains as reported by Winter (1929).
The first estimates of genetic variance in the selected strains were from a half-sib mating design in generation 65 of IHO, ILO, IHP, and ILP (Dudley and Lambert, 1969). Significant genetic variance was found for both oil and protein in all four strains. Genetic variance in IHO was larger than in ILO for both oil and protein with the difference being larger for oil. Similarly the genetic variance in IHP was larger for both oil and protein than in ILP. Predicted gain for one cycle of 20% selection for the selected trait in all four strains was similar to the average gain per generation for the preceding 65 generations.
Genetic variances among S1 lines were estimated for the earliest (generation 69) and latest (generation 98) cycles available of IHP, IHO, and SHO (Dudley and Lambert, 2004). No significant differences in genetic variances between generations 69 and 98 were found for the selected trait in any of the strains tested (Dudley and Lambert, 2004) although significant genetic variance was found in all three strains.
Random Mating
Moreno-Gonzalez et al. (1975) reported estimates of genetic variance obtained from a Design III study of the F2 and random-mated F6 generations of the cross of generations 70 of IHO x ILO. In accordance with quantitative genetic theory, estimates of additive genetic variance in the F6 were significantly lower than in the F2 indicating the build-up of coupling phase linkages for alleles for high oil in IHO and for alleles for low oil in ILO. Estimates of dominance variance were similar in the F2 and F6 suggesting a lack of effect of linkage on dominance variance.
In a similar Design III study of the F2 and F6 generations of the cross of generations 70 of IHP x ILP (Dudley 1994), additive genetic variance for protein was significantly less in the F6 than in the F2 suggesting the build-up of coupling phase linkages for high protein alleles in IHP and for low protein alleles in ILP. The measurement of marker-QTL associations in the F2 and in subsequent random-mated generations was proposed as a method of identifying tight marker-QTL associations.
Regression
Dudley et al. (1974) first reported use of regression of means on cycles of selection to measure rate of progress through generation 70. The experiment was divided into segments based on changes in breeding procedure and the selection environment. Response was measured by regression of means on cycles within segments. Realized heritabilities were calculated for each segment as regression of means on cumulative selection differentials. Results of this process also were reported after 90 (Dudley and Lambert, 1992) and 100 (Dudley and Lambert, 2004) generations. The change per generation for oil in IHO per generation ranged from 2.2 ± 0.7 g kg–1 oil in the first nine generations to 1.6 ± 0.1 g kg–1 in generations 59 to 100 (Dudley and Lambert, 2004). The change per generation in generations 0 to 9 of ILO (–2.1 ± 0.3 g kg–1 oil) was very similar to IHO but from generations 59 to 88 the change was a nonsignificant –0.1 ± 0.4 g kg–1 oil per generation. In contrast, the realized heritability in generations 0 to 9 was 0.50 for both IHO and ILO and in generations 59 to 88 realized heritability was 0.08 for both strains. Inbreeding coefficients in generation 70, calculated based on effective population sizes each generation, were greater than 0.8 for all strains.
Number of Effective Factors
Dudley (1977) evaluated theoretical limits to selection based on additive genetic variance, frequency of alleles favorable for high oil or high protein (q), and number of effective factors. From this study, estimates of 54 effective factors differentiating IHO and ILO for oil and 122 effective factors differentiating IHP and ILP for protein in generation 76 were obtained. Estimates of q were 0.25 for oil and 0.37 for protein in the original Burr's White. Based on these estimates, the observed progress from selection was well within what might be expected with relatively low frequencies of favorable alleles in the original population and a reasonable number of segregating loci. At generation 90, gene frequency estimates for the original population (0.22 for oil and 0.32 for protein) (Dudley and Lambert, 1992) were slightly lower than estimated by Dudley (1977). Because additional progress had been made in IHO and IHP, estimates of effective factors increased to 69 for oil and 173 for protein. As expected, with continued progress in the high direction, estimates of q for oil (0.20) and for protein (0.24) in the original Burr's White were slightly lower after 100 generations (Dudley and Lambert, 2004) than when estimated after 90 generations.
Correlated Response
Each generation a sample of the ears from the strains selected for oil was analyzed for protein and a sample of the strains selected for protein was analyzed for oil (Dudley et al., 1974). From results of the first 70 generations, protein concentration increased in both IHO and ILO from 109 to approximately 150 g kg–1. In the IHP strain, oil did not change whereas in the ILP strain oil concentration decreased from 47 g kg–1 in generation 0 to 31 g kg–1 in generation 70. Replicated trials of selected generations from generation 65 to 100 in IHO showed a significant increase from 148 to 158 g kg–1 protein and a decrease in starch from 435 to 336 g kg–1 as oil increased from 161 to 224 g kg–1 (Dudley and Lambert, 2004). Similar results were obtained in the latest generations of RHO and SHO. In IHP, there was no significant change in oil from generation 67 to 99 but starch decreased from 327 to 234 g kg–1 as protein increased from 228 to 269 g kg–1. Similar results were obtained in RHP and RLP in that oil did not change and starch decreased as protein increased and increased as protein decreased.
Estimates of phenotypic correlations were available for a number of studies of crosses of IHO x ILO and IHP x ILP. In general, correlations of starch with protein were highly negative (ranging from –0.82 to –0.96) in the IHP x ILP cross regardless of random-mated generation (Table 1 ). Correlations of starch with protein in the IHO x ILO cross were lower (varying from –0.21 to –0.73). The correlations of starch with oil were similar in both crosses with a tendency for the RM7 and RM10 generations to have correlations of greater magnitude than for the earlier random-mated generations. The correlations of protein with oil were generally low but positive in all random-mated generations of both crosses. Genetic correlations were similar to phenotypic correlations. These results agree with the results from the observed means in predicting that selection for higher protein will result in decreased starch with little change in oil and selection for higher oil tends to reduce starch and may result in a small increase in protein.
|
QTL Studies
The QTL studies aimed at identification of significant marker-QTL associations in progenies resulting from crosses of divergently selected strains have encompassed a wide range of types and numbers of progenies as well as types and numbers of markers. In most of the studies, classical quantitative genetic information such as means, genetic variances, and heritabilities was available and used to help interpret the data. Emphasis in several studies was on use of random mating to identify tight linkages between markers and QTL. All studies discussed used progenies from the cross of either IHO x ILO or IHP x ILP.
The first study reported in the literature involved 100 F3 progenies from the cross of generation 76 of IHP x ILP and 90 restriction fragment length polymorphism markers (Goldman et al., 1993; Goldman et al., 1994). Twenty-two significant marker QTL associations were found for protein, 19 for starch, and 25 for oil. The number of significant marker-QTL associations for oil (Goldman et al., 1994) was similar to the number found for protein by Goldman et al. (1993) despite the fact that the parents (IHP and ILP) were more divergent for protein than for oil and the genetic variance for protein would have been expected to be higher for protein than for oil.
Random Mating and QTL Identification
The first report comparing marker-QTL associations in non–random-mated and subsequent random-mated generations was by Dudley et al. (2004). Five to seven plants from IHP generation 70 were crossed to five to seven plants from ILP generation 70. The Syn0 (F2) was random-mated four times to produce the Syn4. Two hundred S1 lines each from the Syn0 and the Syn4 generations were evaluated as lines per se and as testcrosses to two inbred testers. Means for starch, protein, and oil of the S1 and testcrosses were similar in the Syn0 and Syn4 for each trait. In accordance with theory and the results of Dudley (1994), genetic variances for protein and starch were significantly less in the Syn4 than in the Syn0 for all progeny types suggesting the build-up of coupling phase linkages for protein and starch.
Dudley et al. (2004) hypothesized that the proportion of significant marker-QTL associations should be higher in the Syn0 than in the Syn4 and that marker-associated effects should be larger in the Syn0 than in the Syn4. For protein, there were 38, 32, and 32 significant marker-associated effects (for the S1, testcross to FR1064, and testcross to FR616 progenies respectively) identified by single marker analysis (SMA) in the Syn0 and 13, 7, and 3 in the Syn4. Similar results were obtained for starch. Because 64 markers were used, the proportions of significant markers were approximately 0.5 in the Syn0 and <0.2 in the Syn4. Thus random mating reduced the proportion of significant marker-associated effects. In addition, for both the S1 and testcross progenies, the mean absolute values of significant marker-associated effects for the Syn4 for protein and starch were approximately half those for significant effects in the Syn0. Unlike the results of Goldman et al. (1993, 1994) the number of significant marker-associated effects for oil was less than the number for protein or starch.
A study similar to that of the IHP x ILP cross was reported by Willmot et al. (2006) in the IHO x ILO cross. Five to seven plants from IHO generation 70 were crossed to five to seven plants from ILO generation 70. One hundred sixty-seven S5 lines from the RM1 and 195 S5 lines from the RM5 and their testcrosses to Illinois Foundation Seeds inbred FR1064 were evaluated. One hundred forty-seven simple sequence repeat markers were used. Unlike the study of Moreno-Gonzalez et al. (1975) there was a significant increase in mean oil concentration from the RM1 to the RM5 for both the lines per se and their testcross progenies. Heritability for oil was approximately 0.90 for both the RM1 and the RM5 for both lines per se and testcrosses. For oil, approximately two-thirds as many marker-associated effects were identified as significant by single marker analysis in the RM5 as were significant in the RM1 for both the lines per se and the testcrosses. Thus, in agreement with the results of Dudley et al. (2004) for protein in the cross of IHP x ILP, random mating reduced the number of significant marker-QTL associations.
Extensive QTL Studies
A much more extensive study of the cross of IHP x ILP generation 70 was reported by Dudley et al. (2007). In this study, phenotypic data were obtained from 500 S2 lines, derived from the RM7 generation, and their testcrosses to a Monsanto company tester. The lines were genotyped for 499 single nucleotide polymorphism (SNP) markers. For the S2 lines, three environments in each of two years were grown while for the testcrosses three environments were used but only one year. Heritabilities for oil, protein, and starch were all above 0.89 for the S2 lines per se and around 0.8 for the testcrosses (Table 2
). Genetic variance for protein was less than reported by Dudley et al. (2004) for the RM1 of the same cross even though the years and number of progenies were greatly different in the two studies. The number of marker-associated effects (measured by SMA) significant at the 0.01 level for protein and starch was approximately 100 for both the per se and testcross progenies. This number is approximately 20% of the total markers used, a proportion quite similar to that found for the Syn4 in the Dudley et al. (2004) study of the same cross using a much smaller number of markers. The number of markers showing significant effects for oil was approximately half the number showing significant effects for protein and starch. This result differs from those of Dudley et al. (2004) and Goldman et al. (1993, 1994). In the Goldman et al. studies, the number of markers significant for oil was similar to the number for protein and in Dudley et al. (2004) the number significant for oil was only 20% of the number significant for protein.
|
Agreement Between IHO x ILO and IHP x ILP
Because of the large number of markers, many of which were closely linked, several markers may be linked to the same QTL. Thus, significant regions were defined as areas where the end of a group of significant markers (based on SMA) was at least 10 cM from the beginning of another region using a Monsanto company composite map. These regions were used to measure agreement in regions identified as significant between the results from the IHO x ILO cross and the IHP x ILP cross (Table 3
). Approximately one-third of the SMA regions identified as significant for protein in the IHP x ILP cross were also identified in the IHO x ILO cross (Table 4
). Similarly, approximately one-third of the regions identified as significant for oil in the IHO x ILO cross were identified as significant in the IHP x ILP cross. Thus divergent selection for protein allowed identification of regions significant for oil not identified in the IHO x ILO cross and divergent selection for oil allowed identification of significant regions for protein not identified in the IHP x ILP cross. For starch, over three-fourths of the regions significant in the per se data for the IHO x ILO cross were also significant in the IHP x ILP cross while over half of the regions significant in the testcross data for IHO x ILO were also significant in the IHP x ILP cross.
|
|
Estimates of Number of QTL
Schon et al. (2004) demonstrated that the number of QTL identified in a given cross increased as the number of progeny tested increases. Because of the large and similar numbers of lines and markers used, the IHP x ILP data reported by Dudley et al. (2007) and the IHO x ILO data reported by Clark et al. (2006) were used for estimating numbers of QTL. In the preceding sections, experiments were compared using the number of markers or regions showing significant effects from SMA. This was done because it was about the only way to relate the results of the studies comparing different random mating generations (Dudley et al., 2004; Willmot et al., 2006) with the extensive studies of the advanced random-mated generations (Laurie et al., 2004; Clark et al., 2006; Dudley et al., 2007). However, because many markers are closely linked, particularly in studies using hundreds of markers, comparisons of numbers of significant markers does not reflect the number of QTL. Thus for estimates of numbers of QTL, a measure of the number of significant chromosomal regions was developed. In the studies of IHO x ILO by Clark et al. (2006) and of IHP x ILP by Dudley et al. (2007) regions were defined by simple interval mapping (SIM) where the end of one region was at least 10 cM from the beginning of the next region.
The number of significant regions defined by SIM for the IHP x ILP RM7 and the IHO x ILO RM10 ranged from 41 to 50 for protein, 54 to 70 for oil, and 42 to 60 for starch (Table 4). The numbers for oil are similar to the number (54) estimated in generation 70 using classical quantitative genetic procedures (Dudley, 1977) although the number (122) estimated for protein by Dudley (1977) was somewhat higher than the numbers estimated here. Several points are of interest. First, the number of significant regions for protein was very similar for the two crosses. Second, the number of significant regions in the testcrosses was similar to the number for the lines per se for all traits. For oil, surprisingly, the number of significant regions measured by SIM was larger in the IHP x ILP cross than in the IHO x ILO cross (Table 4). For starch, the number of significant regions for the IHP x ILP cross was larger than for the IHO x ILO cross.
The finding of relatively similar numbers of significant regions in the IHP x ILP cross and the IHO x ILO cross for both protein and oil are puzzling in that divergent selection for protein in IHP x ILP would be expected to have created parents with more loci having divergent alleles for protein than would divergent selection for oil in the IHO x ILO cross. Likewise, divergent selection for oil would be expected to have created divergence for more loci for oil in the IHO x ILO cross than would divergent selection for protein in the IHP x ILP cross. These expectations are supported by the large divergence in means for protein between IHP and ILP and the relatively small difference in means for oil between IHP and ILP (Table 5 ). Similarly the difference in means for oil between IHO and ILO was large and the difference between those strains for protein was small. The agreement in number of QTL between per se and testcross progenies also is puzzling because the genetic variance in the testcrosses is always considerably less than in the per se progenies (Table 2). However, for all three traits (protein, oil, and starch) the number of significant regions was similar in the per se and testcross progenies (Table 4). In an effort to understand this apparent anomaly, the magnitude of significant effects in the two crosses was compared (Table 2) using SMA data. For oil, the absolute value of significant effects in the IHP x ILP cross was considerably lower than in the IHO x ILO cross. For protein, the absolute value of significant effects in the IHO x ILO cross was lower than in the IHP x ILP cross. In both crosses the absolute value of significant effects is smaller for testcrosses than for lines per se. The question then became how could smaller effects for oil be detected in the IHP x ILP cross than in the IHO x ILO cross and in testcrosses compared to lines per se? A possible answer lies in the fact that the among families within marker genotypes mean square, which is the error term for testing significance of marker-associated effects, includes the genetic variance for the trait being evaluated. For oil, the genetic variance in the IHO x ILO cross is much larger than in the IHP x ILP cross and the average error mean square for testing significance of marker-associated effects is larger (Table 2). Thus, it is possible to detect smaller effects for oil in the IHP x ILP cross. Similarly, the genetic variance for protein is lower in the IHO x ILO cross than in the IHP x ILP cross and thus the error mean square for protein is smaller in the IHO x ILO cross and smaller effects can be detected as significant. Likewise, the genotypic variance in testcrosses is smaller than for lines per se resulting in lower error mean squares for testcrosses than for lines per se.
|
|
Correlated Response
Quantitative trait loci results were used to evaluate relationships among traits by measuring the number of significant SIM regions in common between traits in the IHO x ILO and IHP x ILP crosses (Table 3). In the IHP x ILP cross 40 of 50 SIM regions significant for protein contained significant QTL for starch in the per se progenies and 36 of 41 regions significant for protein in the testcross progenies contained significant QTL for starch (Table 4). Thus, more than 80% of the significant regions for protein were also significant for starch. By contrast, only about one-half of the regions having significant oil QTL had significant QTL for protein. For the IHO x ILO cross, slightly more than half the regions significant for protein in the per se and testcross progenies were significant for starch. For oil, approximately two-thirds of the regions significant for oil in the IHP x ILP cross were also significant for starch whereas in the IHO x ILO cross approximately half of the regions significant for oil were also significant for starch. These results parallel the genetic correlation results.
QTL Summary
Quantitative trait loci studies agree with classical quantitative genetic results in predicting that oil, protein, and starch concentration in the corn kernel are controlled by large numbers of genes, and that selection for protein will result in lower starch concentration, and little change in oil. However, only the QTL studies can lead to location of the QTL involved and to identification of the nature of the genes involved. Comparison of results from large-scale studies of advanced random-mated generations of IHP x ILP and IHO x ILO demonstrated that neither divergence in means of parents nor magnitude of genetic variance was a good predictor of number of QTL detectable as significant.
|
Explanation of Long-term Response |
|---|
|
TOPABSTRACTINTRODUCTIONResultsExplanation of Long-term...ConclusionsREFERENCES |
|---|
1. Large numbers of loci at low gene frequency in the original population. The results presented here, both from "classic" quantitative genetic studies and QTL studies suggest the presence of a large number of segregating QTL affecting oil and protein in the original Burr's White cultivar. The large number of QTL and estimates of low allelic frequency in the original population, support the hypothesis advanced by Dudley (1977) that the long-continued response to selection could be explained by the presence of a large number of QTL with relatively low gene frequency segregating in the base population. The number of QTL estimated is likely a lower limit for the number segregating in the original Burr's White because of the small number of parents (5–7) used in making the crosses studied.
2. Release of epistatic variability. When epistasis exists, selection can generate "new" additive genetic variance rather than depleting it as expected under single locus genetic theory (Goodnight, 2004). In simulation studies, epistasis increased the number of generations during which there was a linear response rather than increasing the rate of response (Goodnight, 2004; Liu et al., 2003). Thus Goodnight (2004) proposed epistasis must be considered as a possible cause of long-continued selection response in the Illinois long-term selection experiment. Eitan and Soller (2004) presented the same concept and called it selection induced genetic variation. They argued that many QTL exhibit epistatic interactions with other QTL or with genetic background. Because selection changes gene frequency at many loci, the genetic background changes. Quantitative trait loci which are neutral in one genetic background may become major QTL as the background changes. Thus new genetic variability is continually released and may account for long-continued selection response. Data supporting this assumption were presented by Carlborg et al. (2006) who showed in broiler chickens that an apparently minor QTL when placed in alternate homozygous backgrounds of other QTL had a major effect in one background and no effect in the alternate background.
Eitan and Soller (2004) suggested negative heterosis in the cross of a selected population to the original population or an earlier version of the population would provide evidence of the presence of additive x additive epistasis. Crosses of IHO x ILO, IHO x RLO, SHO x ILO, SHO x RLO, and RHO x ILO all showed significant negative heterosis for oil (Table 7 ; Dudley et al., 1977). Heterosis was positive for the crosses of IHO x RHO and SHO x RHO. Thus genetic background affected heterosis, supporting the hypothesis that additive x additive epistasis was important. Further evidence comes from the Design III study of Moreno-Gonzalez e t al. (1975) where crosses of both the F2 and F6 of the cross of IHO x ILO back to the parents exhibited negative heterosis for oil (Table 7). This hypothesis is further supported by the presence of significant negative heterosis for protein in the crosses of IHP x RLP and IHP x ILP (Table 8 ; Dudley et al., 1977). Although dominance for low oil or protein might be considered a cause of the negative heterosis for oil or protein it would not explain the positive heterosis in the crosses of IHO x RHO and SHO x RHO. However both sets of results are consistent with the presence of additive x additive epistasis.
|
|
4. Mutation. Walsh (2004) argued that mutation was a necessary assumption to explain the results of the long-term selection experiment. Keightley (2004) reviewed selection experiments in inbred lines and concluded mutational variance was important in selection response. However, neither Walsh nor Keightley considered the effects of epistastic interactions on selection response. Goodnight (2004) suggested gene interaction might magnify the effect of beneficial epistatic mutations.
Summary: Explanation of Response
The available data, from both classical quantitative genetics and QTL studies, show that a large number of QTL were segregating in the original Burr's White. Classical quantitative genetic studies suggest that favorable alleles for high oil or protein were present in low frequency in Burr's White. These results support the concept that long-continued response to selection can be explained by a large number of loci segregating in the original population. The presence of negative heterosis in several crosses involving the selection strains for oil and for protein suggests epistatic effects are important. Change in environment affected response to selection for high protein but not necessarily for oil. Although mutation likely played a role in the long-continued progress, it may be subordinate to the other possible explanations and may not be separable from selection induced variation.
|
Conclusions |
|---|
|
TOPABSTRACTINTRODUCTIONResultsExplanation of Long-term...ConclusionsREFERENCES |
|---|
Taken as a whole, the data support the assumption that progress observed could be the result of a combination of segregation of a large number of genes at low frequency in the original population and selection induced variation resulting from additive x additive epistasis. At the same time, changes in the environment likely affected progress for protein. The possibility that mutation aided selection progress cannot be ruled out.
The contributions of the numerous scientists who maintained continuity in the Illinois long-term selection program over more than 100 yr are gratefully acknowledged. A special thanks to R.J. Lambert, a colleague who worked with the author on this experiment for nearly 40 yr, and Don Roberts, the technician on the project for the last 20 yr. Thanks to G.R. Johnson for numerous stimulating discussions as this paper was being written
Received for publication July 9, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONResultsExplanation of Long-term...ConclusionsREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
