| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Vegetable Crops, Univ. of California, One Shields Ave., Davis, CA 95616 USA
b Dep. of Agronomy, Univ. of Wisconsin, 1575 Linden Dr., Madison, WI 53706 USA
tcosborn{at}facstaff.wisc.edu
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
Abbreviations: cM, centimorgan • WI, winter injury • FG, fall growth • FI, freezing injury • UIL, unifoliate internode length • SDRF, single dose restriction fragment • QTL, quantitative trait locus • RFLP, restriction fragment length polymorphism
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
In the absence of direct measurements of winter hardiness, the winter hardiness potential of genotypes and cultivars has been predicted from related traits that are believed to represent components of winter hardiness. Fall dormancy has traditionally been the primary component used to predict winter hardiness in alfalfa. Short days and cool temperatures trigger the fall dormancy response (Nittler and Gibbs, 1959), which can be scored easily by measuring vertical regrowth following the last fall cutting in the establishment year. Strong positive correlations between fall growth (FG) and winter injury (WI) were reported on the basis of analysis of alfalfa cultivars covering a broad range of fall dormancies (Smith, 1961; Schwab et al., 1996) and of crosses of sensitive and hardy parents (Kohel and Davis, 1960; Brouwer et al., 1998). In contrast, Busbice and Wilsie (1965) found no relationship between FG and WI using F2 progenies, and some recent cultivars have more winter hardiness than expected on the basis of fall dormancy (Barnes and Marten, 1991), indicating that recombination between fall dormancy and winter hardiness may be possible. Since seed yield (Smith, 1961), regrowth after cutting (Busbice and Wilsie, 1965) and yield during the first two production years (Ipson, 1991) were negatively correlated with fall dormancy, the development of non-dormant alfalfa cultivars with good winter hardiness is desirable.
Freezing tolerance is another important component used to predict winter hardiness. Freezing tolerance develops in response to decreasing day length and cold temperatures during autumn (McKenzie et al., 1988), and it can be measured by freezing plants and evaluating frost damage or analyzing cellular leachates. A significant positive relationship between FI and WI has been reported from crosses of sensitive and hardy parents (Kohel and Davis, 1960; Brouwer et al., 1998) while both significant positive (Perry et al., 1987; Brouwer et al., 1998) and non-significant correlations (Kohel and Davis, 1960) have been reported between FI and FG. In a recent study using reciprocal cleft grafts, a graft transmissible factor was identified that affected fall dormancy, but this factor did not affect freezing tolerance, indicating that these two traits may be under different genetic control (Heichel and Henjum, 1990).
Contraction of the hypocotyl and first few internodes also may be related to winter hardiness. This is caused by lateral cell enlargement in young alfalfa seedlings which pulls these first few internodes below the soil surface (Teuber and Brick, 1988). Unifoliate internode length (UIL) of seedlings is an indirect measure of internode contraction, and strong positive relationships between UIL and FG (Schneider et al., 1984; Brouwer et al., 1998), UIL and WI, and UIL and FI (Brouwer et al., 1998) have been reported for crosses of winter-sensitive and winter-hardy parents.
Molecular markers provide a powerful tool for identifying genes controlling complex traits, determining gene action, and obtaining evidence for and against pleiotropy (Paterson et al., 1991; Stuber et al., 1992; Tanksley et al., 1989). For example, Champoux et al. (1995) found that 12 of the 14 chromosomal regions in rice (Oryza sativa L.) containing putative QTLs for root morphology traits also contained QTLs associated with drought avoidance, indicating that selection for root morphology traits may be a effective strategy for improving drought resistance. Pan et al. (1994) identified a marker interval controlling winter hardiness that also contained QTLs for growth habit and heading date in barley (Hordeum vulgare L.); however, unique phenotype combinations in the mapping population suggested that the association between heading date and winter hardiness was due to linkage rather than pleiotropy. Molecular markers were also used to show that separate QTLs controlled vernalization requirement and frost tolerance in wheat (Triticum aestivum L.; Galiba et al., 1995), but QTLs affecting vernalization requirement and freezing tolerance mapped to the same genomic interval in oilseed rape (Brassica napus L.; Teutonico and Osborn, 1995; Teutonico et al., 1995).
Molecular markers have been used to develop linkage maps for diploid (Brummer et al., 1993; Kiss et al., 1993; Echt et al., 1994; Tavoletti et al., 1996) and tetraploid (Brouwer and Osborn, 1999) alfalfa, but these markers have not been used to map QTL in tetraploid alfalfa. The use of molecular markers to map genomic regions affecting fall dormancy, freezing tolerance, and winter hardiness would improve our genetic understanding of these traits and provide insight into the genetic basis for using fall dormancy and freezing tolerance as predictors of winter hardiness. The goal of this research was to identify and compare some of the genomic regions controlling WI and three related traits, FG, FI, and UIL.
|
Materials and methods |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
An RFLP genetic linkage map was created from 82 single dose restriction fragments (SDRFs) in each backcross population (Brouwer and Osborn, 1999). Four cosegregational groups, corresponding to the four homologous chromosomes for each linkage group in the F1, were generated for seven of the eight linkage groups. The two derived from B17 were designated A and B, and the two derived from P13 were designated C and D. Markers on linkage group 7 of the diploid alfalfa maps (Brummer et al, 1993; Echt et al., 1994) were monomorphic and no cosegregation groups corresponding to linkage group 7 were mapped in the two backcrosses. SDRFs were mapped to all four homologues for the remaining 7 linkage groups giving 28 cosegregation groups with average spacing between markers of 12.2 centimorgans (cM), but large gaps were present on cosegregation groups 1B, 2D, 6D, and 8A-D. The composite map including all SDRF loci covered 443 cM on seven linkage groups and had similar genome coverage as previously reported diploid alfalfa maps (Brouwer and Osborn, 1999).
Assessing Plant Phenotypes
Phenotypes were assessed for the same genotypes of the two backcross populations used to create the linkage maps (Brouwer and Osborn, 1999). Unifoliate internode length was measured on all seedlings (5 wk old) in each backcross population as the distance between the cotyledonary node and the unifoliate leaf node. The seedling-derived backcross plants were maintained in the greenhouse, and cuttings for the two replicated field trials were made from two flushes of growth in April and May of 1995 and 1996. Cuttings were used instead of inbred progenies for replicating genotypes to avoid the confounding effects of inbreeding depression, since cuttings faithfully represent seedling-established plants for the traits measured in this study (Brouwer et al., 1998). In the first year of the field trial, cuttings were space-planted at Arlington, WI, on 27 and 28 July 1995 in a randomized complete block design with three replications. Each plot contained 12 plants of one genotype planted 30 cm apart in a row with 60 cm between rows and 1m between the ends of plot rows. This trial included the 202 backcross progenies and two entries each for B17, P13, and the F1. The second year of the field trial was planted at Arlington, WI, on 15 and 16 July 1996 and followed the first year's layout. Since some of the original P13 backcross plants died in the greenhouse, this field trial included 101 B17 backcross progeny, 69 P13 backcross progeny, and two entries each for B17, P13, and the F1. Three dormant, winter hardy check cultivars (Columbia 2000, Saranac, and Ranger) and one non-dormant and non-winter hardy check cultivar (Cuf101) were included in the second year of the field trial.
Fall growth was measured as the height of vertical regrowth in centimeters in early October, FI was measured by absorbance at 265 nm and electrical conductivity (S m-1g-1 x 104) in early November, and winter hardiness was scored in early May as winter survival (percentage of plants scored for FG that were alive in May) and WI (1–5 scale, where 1 is no injury and 5 is dead) in both years of the field trial as described by Brouwer et al. (1998). The only difference was that the measurement of absorbance and electrical conductivity was done on one plant from each replication instead of sacrificing one entire replication. Electrical conductivity and absorbance both measure FI; however, electrical conductivity had a lower coefficient of variation and was highly correlated with absorbance (r = 0.57 in the first year and r = 0.86 in the second year). Thus, electrical conductivity was used as the only measurement of FI and absorbance was dropped from the analysis. Similarly, winter survival and WI both measured winter hardiness and were highly correlated (r = 0.96 in both years). Since WI can account for variation in the health of surviving plants and had a lower coefficient of variation than winter survival, it was chosen as the only measurement of winter hardiness in this study.
Analysis of variance within each backcross population was performed on the plot means by PROC GLM of SAS (SAS Institute, Inc. 1991). Genotypes and years were considered as random effects. Broad sense mean heritabilities of clones were calculated using the mean squares from the SAS GLM procedure as
where
. Phenotypic correlations between the traits were obtained separately for each backcross population and each year using genotype means across replications and were calculated by Microsoft Excel (Microsoft, Tacoma, WA).
Mapping Quantitative Trait Locis
Single factor analyses of variance were computed for each SDRF-trait combination for each year and the 2-yr average following Edwards et al. (1987) from the trait means across replications. The trait values of all individuals having an SDRF fragment were compared to those of all individuals without the fragment by F-tests. Contrasts significant at P < 0.005 were interpreted as strong evidence for linkage between a QTL(s) and an SDRF, and contrasts significant at 0.005 < P < 0.05 were interpreted as weaker evidence for a QTL(s) linked to the SDRF. The lower significance level (P < 0.05) was chosen to control the Type I error at 5% per comparison so that more QTLs with small effects and/or loose linkage to the SDRFs would be detected. Since this was the first comparative mapping study of QTLs for FG, FI, and WI, we felt that controlling Type I errors at the expense of committing a Type II error of declaring a false positive was appropriate.
Multiple regression analyses was performed with data for all 82 SDRFs to identify the best polygenic models for each trait in each backcross population. The best multiple regression model was identified by SAS proc REG (SAS Institute, 1991). Since computations with SAS proc REG become prohibitively large when more than 33 markers are considered, the 82 SDRFs were randomly divided into three subgroups containing 27 or 28 markers each. The best models were first identified by adding markers by the STEPWISE subcommand. All markers in each subgroup that were significantly (P < 0.05) associated with the trait when added to the model were included in a final stepwise procedure to identify the best single multilocus model. The best possible models were also derived by the RSQUARE subcommand of SAS. The 11 markers that detected the largest R2 value from each of the three subgroups (33 markers total) were included in a final selection procedure. Ten models were selected, each of which contained from one to 10 SDRFs with the highest R2 (100 models total). The results of the two model selection methods were compared. For the models explaining FG in the first year field trial, FT in the first year, and WI in the second year, each SDRF entering the STEPWISE model also occurred in most of the 10 RSQUARE models that contained the same number of SDRFs. However, for FG in the second year, FI in the second year, and WI in the first year, some markers entering the STEPWISE model were only found in one or two of the 10 best RSQUARE models containing the same number of SDRFs. In these cases, markers that occurred in a majority of the ten RSQUARE models were used to build a new model by sequentially adding the next most prevalent marker until no new markers added to the model remained significant (P < 0.05) on the basis of the Type III sums of squares. The R2 values of the final models for FI in the second year and WI in the first year were higher than the STEPWISE models, but the final model for FG in the second year had a R2 value slightly lower (0.4%) than the model selected by the STEPWISE subcommand.
Single factor analysis of variance was also performed for marker loci in which two-alleles derived from one parent could be followed in the segregating backcross progenies. Two-allele marker loci calculated for probes that detected two unique SDRFs donated by one parent to the F1 were scored as having 0, 1, or 2 doses of alleles from the non-recurrent parent in the backcross to the other parent. In addition, three pseudo-marker loci having two-alleles were created in the backcross to B17, one on linkage group 1 (pseudo-1) and two on linkage group 5 (pseudo-5a and pseudo-5b), by treating markers linked in repulsion phase as a single marker locus. Genotypes with a recombination event in the region of the pseudo-marker locus were treated as missing data. With these pseudo-marker loci, all homologous linkage groups in both backrosses contained at least one two-allele marker locus. F-tests were used to compare the 0, 1, and 2 dose classes and a P < 0.005 probability level was used to declare the existence of a QTL linked to the two-allele marker loci. At P < 0.005, the 19 comparisons in the B17 backcross and the 16 comparisons in the P13 backcross give approximately a 5% genome wide probability of declaring one false positive (Type II error).
Data from the two-allele marker loci allowed the estimation of gene action in the backcross progeny having 0, 1, or 2 alleles derived from one parent. These effects were estimated by adaptation from Edwards et al. (1987), and are shown below for the B17 backcross:
 |
 |
The dominance gene action was equivalent to the monoplex dominance of Rumbaugh et al. (1988). Estimation of the higher order interactions found in tetraploids, including duplex dominance, trigenic and tetragenic effects, would require the determination of all five genotypic classes at each locus. This is only possible in an F2 population with probes that can follow all four segregating alleles. The di/ai ratio was also computed as a measure of the apparent degree of dominance at each locus for each significant marker-quantitative trait combination (Edwards et al., 1987).
|
Results |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
|
|
Quantitative Trait Loci for Fall Growth
More of the SDRFs were significantly associated with FG in the backcross to B17 (32.3 and 16.4% of the 164 comparisons for the 2 yr at P < 0.05 and P < 0.005, respectively) than in the backcross to P13 (11.0 and 5.2% of the 164 comparisons for the 2 yr at P < 0.05 and P < 0.005 probability levels, respectively). Twenty-two of the 82 SDRFs (26.8%) in the B17 backcross were significantly (P < 0.05) associated with FG in both years compared to only 3 of 82 SDRFs (3.7%) in the P13 backcross. Individual SDRFs accounted for 4.0% to 18.1% of the phenotypic variation for FG in the B17 backcross and 3.9% to 13.7% of the phenotypic variation for FG in the P13 backcross in any one year.
The SDRF multiple regression models for FG in the backcross to B17 included six SDRFs in each year and in the 2-yr average (Table 2). The six SDRFs for 2-yr average FG included SDRFs on 1B, 3A, and 8B, which were significant in both individual years, and SDRFs on 1A, 3B, and 5A, which were significant in only one of the 2 yr. Individual SDRFs in the multiple regression models explained 1.9 to 17.9% of the phenotypic variation in the first year and 2.4 to 9.7% of the phenotypic variation in the second year (Table 2). Marker locus vg2b9 on 1B had the strongest association to FG in the first year but explained only 4.9% of the variation in the second year, while ugac191 on 8B had the weakest association with FG in the first year but the strongest association in the second year. Substituting the presence of a P13 fragment for its absence had a positive effect for all FG QTL in the B17 backcross population (Table 2).
Multiple regression models for the backcross to P13 included three SDRFs associated with FG in the first year, four SDRFs associated with FG in the second year, and three SDRFs associate with 2-yr average FG (Table 3) . Marker locus mtsc35 on cosegregation group 5D was included in all three regression models and was the marker most strongly associated with FG in the first year. Marker locus ugac540 on 8C had the strongest association (P < 0.001) to FG in the second year and the 2-yr average but was not significantly associated with FG in the first year. Significant SDRFs were also detected on 3C for FG in the first year and 2-yr average FG and on 8C for FG in the second year and 2-yr average FG. Likely QTL for FG were associated with an unlinked locus in the first year and on 4D in the second year. The phenotypic variation explained by individual SDRFs ranged from 4.2 to 14.6%. At most of the significantly associated marker loci, substitution of a B17 fragment for its absence had a negative effect on FG; however, positive effects were associated with the B17 fragment for loci on 3C and 6C and an unlinked marker locus (Table 3).
|
|
|
Quantitative Trait Loci for Freezing Injury
More of the SDRFs were significantly associated with freezing injury in the backcross to B17 (20.1 and 16.4% of the 164 comparisons for the 2 yr combined at P < 0.05 and P < 0.005, respectively) than in the backcross to P13 (12.2 and 0% of the 164 comparisons for the 2 yr combined at P < 0.05 and P < 0.005, respectively). Nine of the 82 SDRFs (11.0%) in the B17 backcross were significantly associated with FI in both years while 5 of 82 SDRFs (6.1%) in the P13 backcross were significantly associated with FI in both years. Individual SDRFs accounted for 4.4 to 9.8% of the phenotypic variation for FI in the B17 backcross and for 4.1 to 9.8% of the phenotypic variation for FI in the P13 backcross.
The SDRF multiple regression models for FI in the B17 backcross included three SDRFs in the first year, five in the second year, and five SDRFs for 2-yr average FI (Table 2). Two SDRFs on 1A and 1B were included in all regression models, two SDRFs on 8A and 8B were included in both the second year and 2-yr average FI, and one SDRF on 5B was included in both the first year and 2-yr average FI (Table 2). The multiple regression models indicated only likely associations with QTL (0.005 < P < 0.05) for SDRFs on 1A in the first year and on 5B and 8B for 2-yr average FI; however, the SDRFs on 5B and 8B for 2-yr average FI were each highly significant in one of the 2 yr. Individual SDRFs explained from 3.1 to 9.0% of the phenotypic variation for FI (Table 2). A positive effect of substituting the presence of a P13 fragment for its absence was detected for all significant SDRFs except for the SDRF on 3A detected in the second year (Table 2).
The SDRF multiple regression models for FI in the P13 backcross included two SDRFs in the first year, three SDRFs in the second year, and four SDRFs for 2-yr average FI (Table 3). The strongest association with FI occurred on 5D in all three models, at locus ugac109 in the first year and 2-yr average FI and at locus vg1g9 in the second year. Although ugac109 and vg1g9 are separated by 23.3 cM, all three models detected a negative effect for substituting the presence of a B17 fragment for its absence (Table 2). The remaining SDRFs were associated in only one year (8D in the first year and 3D for 2-yr average FI) or only indicated likely QTL effects (4C and 6C for FI in the second year, 6C and 8D for 2-yr average FI) and substitution of a B17 fragment for its absence had either negative or positive effects at these loci (Table 3).
Three of the two-allele marker loci were significant (P < 0.005) for FI in the backcross to B17. The pseudo-locus on linkage group 1 was significant in both years. This locus was composed of the two SDRFs on 1A and 1B that were associated with FI in the multiple regression models in both individual years (Table 2). The second locus, located on linkage group 8, was significantly associated with FI only in the second year (Table 5). QTLs were also identified on both homologues of linkage group 8 in the multiple regression models in the second year; however, the two most significantly associated marker loci were 32.8 cM apart (Table 2). P13 alleles had positive additive effects and B17 alleles showed partial dominance at all significant marker loci (Tables 4 and 5).
No significant associations between FI and the two-allele marker loci were detected in the P13 backcross population.
Quantitative Trait Loci for Winter Injury
Significant associations between SDRFs and WI were detected for 23.8% and 4.9% of the 164 comparisons for the 2 yr combined at P < 0.05 and P < 0.005 probability levels, respectively, in the backcross to B17. Fifteen of the 82 SDRFs (18.3%) in the backcross to B17 were significantly associated with WI in both years. Individual SDRFs accounted for 5.7 to 17.7% of the phenotypic variation for WI in the first year and 4.1 to 13.2% of the phenotypic variation for WI in the second year. Most genotypes in the backcross to P13 did not survive the winter (Fig. 1), thus, identification of QTL in this backcross was not possible.
The SDRF multiple regression models for the B17 backcross included four SDRFs for WI in the first year, five SDRFs for WI in the second year, and one QTL for 2-yr average (Table 2). SDRFs on 2A, 8A, and 8B were associated with WI in both years, but the SDRF on 8B was the only one associated with 2-yr average WI. Individual SDRFs in the multiple regression model explained 3.7 to 14.7% of the phenotypic variation for WI in 1996 and 2.2 to 9.6% of the phenotypic variation for WI in 1996–1997. A positive effect of substituting a P13 fragment for its absence was detected for all significant SDRFs (Table 2).
Two-allele marker loci were significantly (P < 0.005) associated with WI on linkage groups 5 and 8 in both individual years. Ugac540 on linkage group 8 was significant in each year (Table 4) while ugac191, 32.8 cM from ugac540 on linkage group 8, was only detected in 1996–1997 (Table 5) . This region on linkage group 8 also was significantly associated with WI in the multiple regression models. A pseudo-locus on linkage group 5, pseudo5b, was significant in both 1995–1996 and 1996–1997. The SDRF models did not identify any QTLs on linkage group 5 in 1995–1996, but two SDRFs were identified by the multiple regression models on 5B in the second year. P13 alleles had positive additive effects on WI for all significant marker loci. The P13 alleles at pseudo5b and ugac191 and the B17 allele at ugac540 had slight dominance effects (Tables 4 and 5). The small d/a ratios in Tables 4 and 5 indicated that additive gene action was most important for WI.
Quantitative Trait Loci for Unifoliate Internode Length
Although no SDRFs were significantly associated with UIL in the B17 backcross, 12.2 and 0% of the 82 comparisons in the P13 backcross were significant at P < 0.05 and P < 0.005, respectively. Individual SDRFs accounted for 4.0 to 7.4% of the phenotypic variation for UIL in the P13 backcross.
The SDRF multiple regression model for UIL in the P13 backcross included two SDRFs, both detected by probe UGAC85, on linkage group 3. Negative effects of substituting the presence of a B17 fragment for its absence were detected for both SDRFs.
No comparisons between two-allele loci and UIL were significant in either backcross.
Correspondence between QTLs Affecting FG, FI, and WI
In the backcross to B17, some of the SDRFs associated with each trait in the multiple regression models occurred on homologous cosegregation groups. For FG, SDRFs were identified on both homologues of linkage group 1 in the first year, and in the 2-yr average (Table 2). Two other sets of SDRFs associated with FG were identified on 3A and 3B in the first year and on 8A and 8B in the second year. Similar pairing of SDRFs on homologous cosegregational groups occurred on linkage groups 1 and 8 for FI and on linkage group 8 for WI. In the P13 backcross, both SDRFs detected by probe UWG085 on 3C and 3D were associated with UIL.
Some of the genomic regions analyzed in the B17 backcross were significantly associated with more than one trait. For example, SDRFs on 1A and 1B, separated by only 15.6 cM, were significantly associated with FG and FI, and the presence of the SDRFs (P13 alleles) was associated with larger trait values at each locus (Table 2). A two-allele marker locus in the same region also was associated with FG and FI in both years (Tables 4 and 5). SDRFs significantly associated with all three traits were identified on 8A and 8B. The presence of the SDRFs (P13 alleles) was associated with greater FG, FI and WI at both loci (Table 2), although two-allele models indicated differences in gene action at each locus (dominance of the P13 allele for ugac191 and dominance of the B17 allele for uga540; Tables 4 and 5). In the P13 backcross, an SDRF on 8D (uwg170) was associated with FI in the first year and an SDRF on 8C (ugac540) was associated with FG in the second year. SDRFs that accounted for smaller effects were identified for all traits (except UIL) in at least one year on homologues of linkage group 5 in both backcrosses, and SDRFs for FG and FI were identified in at least one year on homologues of linkage group 3 in both backcrosses. SDRFs associated only with WI were found on 2A and 4B in the B17 backcross.
|
Discussion |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
The B17 backcross population had higher heritabilities for all traits and more QTLs were detected with larger effects than in the P13 population. Positive effects of substituting the presence of a P13 allele for its absence were identified for all traits in this population, in accordance with the P13 parent contributing alleles for greater FG, FI, and WI. QTLs identified in the P13 backcross population were often located in the same regions as QTLs in the B17 backcross. However, the additive effect of substituting a B17 allele for its absence was not always negative, as expected. Most genotypes in the P13 backcross had the erect growth habit typical of non-dormant alfalfa. Variation for FG in the P13 backcross appeared to be associated more with plant vigor than with a fall dormancy response and may be due to heterosis for plant height. If this is true, the SDRFs associated with FG that detected a positive response for substituting a B17 fragment for its absence may actually be associated with QTL for plant vigor. No QTL for WI were mapped in the P13 backcross because no plants survived the winter. Thus, comparisons of genomic regions affecting traits in the P13 backcross were either not meaningful (FG effects due to vigor) or not possible (lack of WI data).
Two studies that compared the observed and expected means of F2 and backcross populations of alfalfa found partial dominance for conductivity (FI) and additive gene action for FG (Kohel and Davis, 1960; Perry et al., 1987). Knipe and Stockton (1977) also found predominately additive inheritance of FG based on comparison of F1 population means to midparent expectations. In our study, comparisons between the observed and expected means of the backcross populations suggested partially dominant gene action of P13 alleles for FG and of B17 alleles for FI, and essentially additive gene action for WI (data not shown). These results are in agreement with those based on the d/a ratios at specific two-allele marker loci (Tables 4 and 5). Most loci associated with FG and FI showed partial dominance for P13 and B17 alleles, respectively, and although partial dominance was detected for WI, the relatively small d/a ratios for WI emphasize the importance of additive gene action for this trait.
The magnitudes of some QTL effects varied greatly between years, suggesting the presence of QTL by environment interactions. In the B17 backcross, most of the QTL effects that were highly significant in one year were considerably less important in the other year. This variation among QTL effects over years was surprising, considering the high correlation between years for FD and WI in this study (Table 1) and in studies by Schwab et al. (1996) which reported correlations between years of r = 0.94 for FG and r = 0.83-0.98 for WI in Minnesota. However, the expression of FG, FI, and WI are all influenced by variable environmental factors. Fall growth is a response to decreasing photoperiod and cool temperatures, FI can be reduced by cold acclimation, and WI is affected by environmental stresses such as ice sheeting and extreme cold which do not occur every winter (McKenzie et al., 1988). Year-to-year variation in the timing of environmental factors may cause variation in the expression of QTL alleles, and this variation in expression could explain some of the conflicting reports of correlations between WI and FG and FI described in the Introduction. However, the genomic regions that we found to have the largest effects in one year also had some significant effect in the other year and in the 2-yr average , suggesting that although the magnitude of the effects varied, some QTLs contributed to the variation every year.
The multiple regression models in the B17 backcross identified several genomic regions associated with FD, FI, and WI. In many of these regions, putative QTLs were identified by the same marker locus or linked marker loci on each homologue. This could be due to the presence of a single QTL with effective alleles on each homologue; although multiple QTLs in these regions can not be ruled out. On linkage groups 5 and 8, marker loci were associated with all three traits in at least one year, and together they accounted for more than half of the variation explained by the multiple regression models for WI in the first year and for FG, FI, and WI in the second year. Consistency across traits in the type of gene action at locus ugac540 (partial dominance of the P13 allele) on linkage group 8A strongly suggested that pleiotropy was the cause for the association among the three traits at this locus. Although ugac540 also segregated on 8B, it was not included in the multiple regression model. Another SDRF on 8B, 32.8 cM from ugac540, was include in the model. This locus, ugac191, had a different type of gene action (partial dominanace of the B17 allele), providing evidence that linkage group 8 contained at least two distinct QTL. Only a subset of traits were associated with marker loci on linkage groups 1 and 3 (FG and FI) and on linkage groups 2 and 4 (WI). Thus, although genomic regions controlling all traits were identified, regions that influence only FG and FI or WI also exist.
The genetic analysis of winter hardiness and its component traits indicates that there is a genetic basis for using predictor traits in the absence of winter hardiness data. Genomic regions controlling the FG and FI also accounted for as much as half of the genetic variance for WI in the QTL models. However, we also found regions associated with FG and FI that were not associated with WI, and visa versa. Thus, indirect selection to increase winter hardiness using FG and FI will not be effective for all loci controlling WI. The additive gene action and high heritability found for WI in this study suggests that direct selection in winters with differentiation between genotypes will be the most reliable way to improve winter hardiness.
Fall dormancy has negative effects on yield and regrowth after cutting (Busbice and Wilsie, 1965) and decreases seed yield (Smith, 1961). Our results, together with reports of variable FG and WI correlations and the release of cultivars with more winter hardiness than expected based on FG scores (see Introduction), indicate that winter hardiness could be improved without increasing fall dormancy. Molecular markers may be useful for this, especially in conjunction with phenotypic selection (Hospital and Charcosett, 1997; Romagosa et al., 1999). For example, markers ugac482 and vg0c5 on chromosome 2 and ugac671 on chromosome 4 could be used to introgress B17 alleles into non-dormant germplasm to improve winter hardiness without affecting FG. Employing several dormant donor genotypes would help to prevent inbreeding depression; however, QTL effects for WI may differ in germplasm not related to the parents used in this study. Further studies are needed to determine the effects of different alleles from these regions, and to determine their effects in different genetic backgrounds.Barnes Martin 1991; Perry Larson 1974
|
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
Received for publication October 15, 1999.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMaterials and methodsResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. Herrmann, B. Boller, B. Studer, F. Widmer, and R. Kolliker Improving Persistence in Red Clover: Insights from QTL Analysis and Comparative Phenotypic Evaluation Crop Sci., January 16, 2008; 48(1): 269 - 277. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. J. Ariss and G. J. Vandemark Assessment of Genetic Diversity among Nondormant and Semidormant Alfalfa Populations Using Sequence-Related Amplified Polymorphisms Crop Sci., November 7, 2007; 47(6): 2274 - 2284. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. G. Robins, D. Luth, T. A. Campbell, G. R. Bauchan, C. He, D. R. Viands, J. L. Hansen, and E. C. Brummer Genetic Mapping of Biomass Production in Tetraploid Alfalfa Crop Sci., January 22, 2007; 47(1): 1 - 10. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Cao, B. A. Craig, and R. W. Doerge A Model Selection-Based Interval-Mapping Method for Autopolyploids Genetics, April 1, 2005; 169(4): 2371 - 2382. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A. Weishaar, E. C. Brummer, J. J. Volenec, K. J. Moore, and S. Cunningham Improving Winter Hardiness in Nondormant Alfalfa Germplasm Crop Sci., January 1, 2005; 45(1): 60 - 65. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Cao, T. C. Osborn, and R. W. Doerge Correct Estimation of Preferential Chromosome Pairing in Autotetraploids Genome Res., March 1, 2004; 14(3): 459 - 462. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Kahraman, I. Kusmenoglu, N. Aydin, A. Aydogan, W. Erskine, and F. J. Muehlbauer Genetics of Winter Hardiness in 10 Lentil Recombinant Inbred Line Populations Crop Sci., January 1, 2004; 44(1): 5 - 12. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Kahraman, I. Kusmenoglu, N. Aydin, A. Aydogan, W. Erskine, and F. J. Muehlbauer QTL Mapping of Winter Hardiness Genes in Lentil Crop Sci., January 1, 2004; 44(1): 13 - 22. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Wu, C.-X. Ma, and G. Casella A Bivalent Polyploid Model for Mapping Quantitative Trait Loci in Outcrossing Tetraploids Genetics, January 1, 2004; 166(1): 581 - 595. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C.-X. Ma, G. Casella, Z.-J. Shen, T. C. Osborn, and R. Wu A Unified Framework for Mapping Quantitative Trait Loci in Bivalent Tetraploids Using Single-dose Restriction Fragments: A Case Study from Alfalfa Genome Res., December 1, 2002; 12(12): 1974 - 1981. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
