| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a former post-doc research assoc., Univ. of Illinois, Urbana, IL 61801
b Dep. of Crop Sciences, Univ. of Illinois, Urbana, IL 61801
c USDA-ARS, Soybean/Maize Germplasm, Pathology, and Genetics Research Unit, Dep. of Crop Sciences, Univ. of Illinois, Urbana, IL 61801
d Dep. of Horticulture and Crop Science, The Ohio State Univ., Columbus, OH 43210
e Dep. of Agronomy, Purdue Univ., West Lafayette, IN 47907
f Dep. of Plant Pathology, Univ. of Wisconsin, Madison, WI 53706
* Corresponding author (psguzman{at}monganto.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abbreviations: BC, backcross • cM, centimorgan • LG(s), linkage group(s) • LOD, likelihood of odds • PCR, polymerase chain reaction • PI, plant introduction • QTL, quantitative trait loci/locus • RFLP, restriction fragment length polymorphism • RI, recombinant inbred • RP, recurrent parents • SSR, simple sequence repeat
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Exotic germplasm has long been tapped to broaden the soybean genetic base for sustained genetic improvement (Thorne and Fehr, 1970), but the utilization of exotic germplasm is hampered by the presence of unfavorable genes tightly linked with the beneficial genes (Concibido et al., 2003), and by the high frequency of deleterious alleles in much of the germplasm. However, molecular marker technology has made it possible to localize and select useful genes, and to discriminate against unfavorable genomic regions resulting in greater interest in the use of exotic germplasm in soybean breeding programs. Orf et al. (2004) stated that "an important use of markers in the coming decade will likely be the mining of new genetic diversity from soybean germplasm collections and collections of related species." The identification of high yielding soybean lines with acceptable agronomic traits derived from soybean PIs indicates the potential of mapping genes that increase yield from exotic germplasm (Thompson and Nelson, 1998; Brown-Guedira et al., 2004; Warburton et al., 2004).
Studies on QTL mapping of soybean yield genes from exotic germplasm are limited. Orf et al. (1999a), Specht et al. (2001), Kabelka et al. (2004), and Smalley et al. (2004) reported QTL alleles from soybean PIs that increase yield. Concibido et al. (2003) found a yield QTL from a Glyince soja (Siebold and Zucc.) accession that increased yield up to 9%. These studies demonstrated the potential of PIs as sources of yield enhancing alleles. More yield QTL studies need to be conducted using different PIs as parents to discover additional genes that could be introgressed into elite U.S. soybean cultivars. The objectives of our study were to map yield QTL in three backcross populations that each have different PIs as donor parents, and to determine yield QTL x environment interactions and QTL epistatic interactions.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
All of the exotic accessions used as parents in developing these mapping populations were selected based on data collected in a cooperative yield testing project initiated by Clark Jennings of Pioneer Hi-Bred International in 1978, and jointly organized by Dr. Jennings and Randall Nelson (USDA-ARS) in subsequent years. This initial series of PI tests was discontinued after 1986. Between 1978 and 1986, approximately 40 public and private soybean breeders in the U.S. and Canada participated in the testing of over 2000 soybean introductions. The entries were selected based on general evaluation data of accessions in the USDA Northern Soybean Germplasm Collection collected between 1964 and 1983. Lawrence (Bernard et al., 1988), and Kenwood (Cianzio et al., 1990) were selected as parents because they were recent releases at the time the first crosses were made. A near isogenic line of ‘Beeson’ (Probst et al., 1969) with genetic male sterility was used as a parent because at that time, it was intended to develop populations segregating for male sterility to facilitate recurrent selection.
In 1979, L74L-125 (later released as Lawrence) was crossed to PI 68658, introduced from northeast China in 1926. The F1 plants were grown in Puerto Rico during the winter, and F2 plants were harvested at Urbana, IL in 1980. Single plants were harvested from the best F2 families in 1981, and LG82–8224 was harvested as an F4 line in 1982. Tests at four locations in Illinois in 1983 and 1984 showed that LG82–8224 was equal to ‘Sparks’ (Nickell et al., 1983) in yield, and 1 d earlier in maturity (unpublished data, 1983 and 1984). In 1985, LG82–8224 was backcrossed to Lawrence, the F1 plants were grown in Puerto Rico the following winter, and the F2 population was grown at Urbana in 1986. An early generation testing procedure was used in which F2 families were yield tested in unreplicated, one row plots in 1987, and in replicated, bordered plots in 1988. In 1989, single plants were selected from the highest yielding populations identified in 1988, and tested in unreplicated, one-row plots in 1990. Selected F6 lines were tested in replicated, bordered plots in subsequent years. Following 2 yr of testing, LG90–4931 was selected as the BC1 line with a yield greater than that of Lawrence (unpublished data, 1991 and 1992). LG90–4931 was backcrossed to Lawrence in the greenhouse during the winter of 1991–92. The resulting population was again tested in an early generation testing procedure as described previously. In 1997 and 1998 in tests at one location with two replications per test, LG96–6607 yielded 17% more than Lawrence with no difference in maturity (unpublished data, 1997 and 1998). LG96–6607 was backcrossed to Lawrence in the summer of 1998, and the F1 plants were grown the following summer. Single plants were harvested from the F2 population in 2000, and 94 F3 plant rows were harvested in 2001. These single plant rows were used as the entries in the Lawrence-derived mapping population.
The initial cross of the Beeson population was made between L75–0587 and PI 407720, introduced from Jilin, China in 1974. L75–0587 (Bernard et al., 1991) is BC5 near isogenic line of Beeson with the ms2 allele from T259 (Nelson and Bernard, 1991). LG84–7963 was a male-fertile F3 selection that was yield tested in 1985 and 1986. LG84–7963 was backcrossed to Beeson 80 (Wilcox et al., 1980) in the greenhouse during the winter of 1986–87. Beeson 80 is a BC7–derived line that differs from Beeson by the Rps1-c allele from Arksoy. The F1 plants were grown at Urbana in 1987. The F2 selections were made in 1988, and evaluated using the early generation testing procedure as previously described. LG92–1143 was an F6 selection that was tested at 8 locations (2 replications per location) over 3 yr, and exceeded the yield of Beeson 80 by 17% with no difference in maturity (unpublished data, 1997 and 1998). LG92–1143 was backcrossed to Beeson 80 in 1998, and the F1 plants were grown in Puerto Rico the following winter. The F2 population was grown at Urbana in 1999, and the F3 and F4 generations were grown in the greenhouse the following winter. The F5 generation was planted in the field in late June and early July, 2000. A frost the first week in September left most plants with fewer than 10 viable seeds. The seeds from all plants were sent to Los Andes, Chile during the winter of 2000–2001 for an increase of the F6 lines, and the resulting harvest was planted in Urbana in 2001 for a final seed increase.
PI 391583, Jilin 10 introduced from China in 1974, was crossed in the summer of 1980, to PI 297544, the Russian cultivar Primorskaja 529, introduced from Hungary in 1964. The F1 plants were grown the following summer. Single plants harvested from the F2 population in 1982 were entered into an early generation testing procedure as previously described. LG87–1606 is an F6 selection that was tested from 1988 through 1990. In 1989, LG87–1606 was crossed to Kenwood and the F1 plants were grown in Puerto Rico the following winter. The F2 plants harvested in 1990 were entered into the early generation testing program. LG94–1713 is an F6 selection that was tested in one location for 3 yr (2 replications per test) and yielded 9% more than Kenwood with no difference in maturity (unpublished data). LG94–1713 was backcrossed to Kenwood in 1998, and the subsequent population was advanced in the same manner as the final backcross in the Beeson population previously described.
Phenotypic Evaluation
The Kenwood population contained 74 BC1 F5-derived lines, the Beeson 80 population consisted of 68 BC2 F5-derived lines, and the Lawrence population had 94 BC3 F2-derived lines. The populations were each evaluated as separate experiments. The recurrent parents were included in the tests, plus two check cultivars in the Beeson 80 and Kenwood experiments, and three checks in the Lawrence experiment. The Beeson 80 and Kenwood experiments were evaluated in 2003 near Urbana, Bellfower, and Dekalb, IL, and Columbus, OH. In 2004, they were evaluated at the three Illinois locations and at West Lafayette, IN. The Lawrence experiment was evaluated near Urbana, Ivesdale, and Hume, IL, and Columbus, OH, in 2003, and at the three Illinois locations in 2004. Each location-year combination was considered a separate environment, for a total of eight test environments for Beeson 80 and Kenwood experiments, and seven for the Lawrence experiment. A randomized complete block design (RCBD) with two replications was used at each environment. Genotypes were grown in four-row plots, with row lengths of 3 m and a 0.61 to 0.76 m row spacing, depending on the location. Agronomic trait data were obtained from the center two rows of plots. These traits were (i) grain yield (Mg ha–1) adjusted to 130 g kg–1 moisture, (ii) days to maturity recorded as the number of days after planting when about 95% of the pods had reached mature pod color (R8; Fehr et al., 1971), (iii) lodging, scored as 1–5 at maturity with 1 representing all plants erect and 5 all plants prostrate, and (iv) plant height (cm), measured as the distance from the ground to the top node of the main stem at maturity. Plots were not end-trimmed before harvest.
Simple Sequence Repeat Marker Analysis
The populations were tested with SSR markers using DNA from leaf tissue collected from eight greenhouse-grown seedlings of each genotype used in the trials. The DNA was extracted from the leaf tissue according to Kabelka et al. (2005). Polymerase chain reaction (PCR) products were obtained using the protocol described by Cregan and Quigley (1997), with non-labeled and fluorescently-labeled SSR primers. The SSR markers were developed by Dr. Perry B. Cregan (USDA-ARS, Beltsville, MD). Non-labeled PCR products were separated through non-denaturing polyacrlylamide gel electrophoresis (PAGE) according to Wang et al. (2003). An ABI Prism 377 Genetic Analyzer (Applied Biosystems, Foster City, CA) was used to analyze the fluorescently-labeled PCR products. DNA from the parents of the mapping populations was first screened against 602 SSR markers covering the 20 chromosomes of the soybean genome. The populations were tested with polymorphic markers, which included 45 in the Beeson 80 population, 84 in the Kenwood population and 30 in the Lawrence population.
Evaluation of Beeson 80 BC Lines for Resistance to Brown Stem Rot (BSR)
Six lines in the Beeson 80 population were tested to determine if they were segregating for a BSR resistance gene in the region on linkage group (LG) J where all BSR resistance genes have been mapped (Lewers et al., 1999; Bachman et al., 2001; Patzoldt et al., 2005). This was done by inoculating three high yielding lines homozygous for the Satt547 allele from the donor parent, and three low yielding lines homozygous for the Satt547 allele from Beeson 80 with Phialophora gregata (Allington and Chamberlain). Also included in the test were BSR resistant cultivars BSR101 (Rbs 1 and Rbs 3; Hanson et al., 1988; Willmot and Nickell, 1989) and Dwight (Nickell et al., 1998), and BSR-susceptible cultivar Corsoy 79 (Bernard and Cremeens, 1988). The tests were done in two greenhouse experiments from 3 Jan. 2005 to 1 Apr. 2005 at the Dep. of Plant Pathology, Univ. of Wisconsin-Madison. Spore suspensions of P. gregata, genotype A isolates IN-6, F5–3, and Fulton-OH, were prepared as previously described (Hughes et al., 2002). Spore concentration of each isolate was determined individually with a hemacytometer (Bright-Line Hemacytometer, Hausser Scientific, Horsham, PA) and concentrations were adjusted to 1 x 107 spores mL–1 before mixing.
Seeds of each soybean genotype were germinated in 15 cm diameter plastic pots containing Scott's (Marysville, OH) Metro-Mix (Experiment 1) or a 1:1 mixture of Scott's Metro-Mix and Fafard (Agawam, MA) Peat Moss (Experiment 2). Each experimental unit was a pot with approximately four seedlings and the experiments were replicated four times. The seedlings were inoculated between the VC–V1 stages (first tri-foliate leaf open but not fully expanded) (Fehr et al., 1971) using a stem-injection method. A 10 mL syringe containing a spore suspension of P. gregata, and fitted to an 18-gauge needle was used to pierce the hypocotyl below the soil surface at the root-stem interface. The needle was inserted half way through the hypocotyl and approximately 200–500 µL of inoculum was directly injected into the vascular and pith tissues. The site of inoculation was then covered with the surrounding potting mix. The center plant of each pot was mock inoculated in the same manner with sterile water to serve as a control.
Seven days following inoculation, pots were arranged in a Completely Randomized Design and fertilized with approximately 8 g per pot of Osmocote 18–6–12 (The Scotts Co., Marysville, OH). Six to seven wk following inoculation, individual plants were rated for BSR symptom development and disease severity by determining the percentage of nodes with leaves expressing BSR symptoms.
Statistical Analysis
Individual environments were analyzed using nearest neighbor analysis (NNA) (Papadakis, 1937) with Agrobase Generation II software (Agronomix Software Inc., Winnipeg, MB, Canada). Adjusted entry means obtained from the NNA were used in the calculation of combined ANOVA over environments, in the QTL analysis in 2003 and 2004, and across all environments. Variance components were estimated by treating the lines in each population, and the replications, and environments as random effects. Heritability and exact 95% confidence intervals (Knapp et al., 1985) were computed on an entry mean basis.
Linkage analysis was done with JOINMAP 3.0 (Kyazma B.V., P.O. Box 182, 6700 AD Wageningen, Netherlands) (Van Ooijen and Voorrips, 2001) using the Kosambi mapping function at a LOD grouping threshold of 3.0. Single-marker analysis was done using one-way ANOVA with PROC GLM in SAS (SAS, 1999). Interval-mapping (IM) and composite interval mapping (CIM) were conducted using MapQTL 4.0 (Van Ooijen et al., 2002). The permutation test option in MapQTL 4.0 was employed to determine the P = 0.05 genome-wide significance level for declaring QTL significant. However, a comparison-wise P < 0.05 was eventually used to declare a putative QTL significant. Composite interval mapping was not done in the Beeson 80 and Lawrence experiments because of the limited number of markers segregating in both experiments.
The proportion of the variance (R2) explained by the QTL and the additive (a) effects were estimated by MapQTL 4.0 at the QTL peaks in the IM and CIM. The total phenotypic variance explained by two or more QTL for a given trait was determined using a multifactor ANOVA that included all significant QTL. The QTL x E interaction for yield was analyzed as a split-plot, with marker as the main plot and environment x marker as the sub-plot (Utz and Melchinger, 1996) using PROC GLM in SAS (SAS, 1999). A QTL x E interaction was declared significant when the P-value was less than 0.05. Digenic epistasis between all pairs of loci was evaluated for yield only with two-locus ANOVA using EPISTACY (Holland, 1998). An epistatic interaction was declared significant when the P value was less than 0.001. The BSR disease ratings were analyzed using PROC GLM in SAS.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
There were significant (P < 0.01) differences between 2003 and 2004 means for all traits except plant height in the Beeson 80 population (Table 1). Differences between 2003 and 2004 means for all traits were significant (P < 0.05) in the Kenwood population. There were significant (P < 0.01) differences between years for yield and days to maturity, but there were no significant differences for lodging and plant height in the Lawrence population. In general, the 2004 environments had more optimal rainfall than the 2003 environments, resulting in greater yields in 2004 than in 2003.
|
|
QTL Analysis
Thirteen putative yield QTL were detected across the three populations (Table 3). Three yield QTL were detected in the Beeson 80 population, while five each were detected in the Kenwood and Lawrence populations. Percent phenotypic variance for yield explained by the individual QTL in the Beeson 80 population ranged from 10 to 51%. Based on the mean over all environments, the three QTL for yield in the Beeson 80 population collectively explained 53% of the phenotypic variance for yield.
|
All of the five QTL for yield in the Kenwood population were detected in the analysis of means across environments, and three of the five were detected in both years. The other two QTL were not detected in 2003. Individually, the phenotypic variation for yield explained by the QTL in the Kenwood population ranged from 5 to 23%. Collectively, the five yield QTL in the Kenwood population accounted for 58% of the phenotypic variance for yield on the basis of means across environments.
Of the five QTL detected in the Lawrence population, three were detected in the analysis of means across environments, and two were identified in 2003 only. The QTL identified on LG A1 was the only one detected in both years and across environments. The phenotypic variation explained by the individual QTL in the Lawrence population ranged from 8 to 26%. The three QTL detected in the means across environments collectively explained 30% of the phenotypic variance for yield.
The yield-increasing allele at eight of the thirteen yield QTL detected in our study were inherited from the PI parents. Four of the eight positive alleles from the PIs were detected in both years and in the mean across environments. In general, the magnitude of the effects of the positive alleles inherited from the PIs were comparable with those derived from the recurrent parents.
One maturity QTL was detected in the Lawrence population, three in the Beeson 80 population and four in the Kenwood population (Table 4). Five maturity QTL were detected in 2003 and 2004, and in the mean across environments. Based on the means across environments, the maturity QTL collectively explained 36% of the phenotypic variance in the Beeson 80 population and 56% in the Kenwood population. The maturity QTL on LGs C2 and O mapped to the same region as the yield QTL in the Kenwood population and in both cases, the allele for later maturity was also associated with greater yield. The maturity QTL detected by Satt556 on LG B2 in 2003 in the Kenwood population was in the same region where a QTL for yield was detected in the Lawrence population. The maturity QTL in the Beeson 80 population and Lawrence populations did not map to regions containing yield QTL. The effect of substituting a Kenwood allele for a PI allele on LG C2 extended maturity to 3.0–3.5 d. Five of the eight QTL alleles for later maturity were derived from the PI parents.
|
Plant height QTL on LGs J and M were identified in the Beeson 80 population, on LG C2 in the Kenwood population, and on LG G in the Lawrence population. Three plant height QTL were detected in 2003, 2004 and in the means across environments. Collectively, 31% of the phenotypic variance for plant height in the Beeson 80 population was explained by the two QTL based on means across environments. The plant height QTL detected by Satt547 on LG J in 2004 and in the mean across environments was also associated with increased yield in the Beeson 80 population. The plant height QTL on LG C2 in the Kenwood population mapped to the same region as the yield and maturity QTL. The effect of substituting an allele of the recurrent parent with the allele of the PI parent on LG C2 increased plant height from 2.7 to 4.1 cm. The QTL allele from the PI increased plant height from 5.0 to 8.0 cm in the Kenwood population. In the Lawrence population, the QTL for plant height on LG G mapped to the same region as the QTL for maturity and lodging. In our study, the QTL alleles for greater plant height on LGs C2 and J were derived from the PI parents while the QTL alleles for taller plants on LGs G and M were inherited from the recurrent parents.
Significant (P < 0.05) QTL x E interactions were detected for six yield QTL across all environments (Table 5). The yield QTL detected by Satt547 in the Beeson 80 population and by Satt557 in the Kenwood population showed significant QTL x E interaction in both years and across all environments. Significant QTL x E interactions for the QTL detected by Satt313 on LG L in the Kenwood population were exhibited in 2004 and across all environments, but not among environments in 2003. Significant QTL x E interaction for the yield QTL marked by Satt622 on LG J in the Lawrence population was only detected in the analysis across all environments.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
= 0.05). When the genome wide threshold of P < 0.05 was used, five were significant for yield. We are aware that the low threshold could result in false QTL declaration, but we are more concerned about avoiding a Type II error. The high heritability for yield obtained in our study would reduce the false discovery rate (Bernardo, 2004), but there is still a need to confirm the validity of all QTL. All of the thirteen yield QTL detected in the three populations mapped to regions where yield QTL were previously reported (Fig. 1 ). Yuan et al. (2002), Kabelka et al. (2004), and Smalley et al. (2004) reported yield QTL within 4 cM of the LG K yield QTL we mapped in the Beeson 80 population. Although Kabelka et al. (2004) reported that the yield-increasing QTL allele on LG K originated from the PI parents, our results and that of Yuan et al. (2002) and Smalley et al. (2004) suggest that a yield-improving allele is also present in the commercial gene pool.
|
A third yield QTL in the Beeson 80 population was mapped to LG J in the vicinity of Satt215. Smalley et al. (2004) also identified a yield QTL on LG J in the same region marked by Satt529 in the AP10 soybean population, which is 100% PI in origin but had been subjected to four cycles of recurrent selection for yield. Satt215 and Satt529 map within 4 cM of one another, and they map within 26 cM from Satt547 (Song et al., 2004). The yield QTL on LG J in the Lawrence population may be allelic with the QTL marked by Satt215 in the Beeson 80 population since Satt215, Satt414, and Satt622 on LG J are within 7 cM of each other. In both populations, the yield increasing allele was from the PI parent.
Yuan et al. (2002) identified a yield QTL in the cross Essex x Forrest that was located in the same region where we mapped a QTL on LG C1 in the Kenwood population. Yuan et al. (2002) reported that the yield QTL associated with Satt294 was detected in one of four environments, while in our study, the yield QTL was detected across four environments in 2004, and across all environments. A yield QTL associated with Satt578 on LG C1 was also identified by Smalley et al. (2004) in the AP10 soybean population. Orf et al. (1999a) identified a QTL for seed weight, a yield component, on LG C1 that was associated with the RFLP marker L192_1 in the cross between the soybean PI Minsoy and the U.S. cultivar Archer. Satt294, Satt578, and L192_1 map 2, 11, 3 cM, respectively, from Satt399 on the integrated map (Song et al., 2004). The proximity of these markers indicates that the same yield QTL may have been detected in all studies. In all studies reporting the LG C1 yield QTL, the yield-increasing allele was from an elite parent.
Orf et al. (1999a), Specht et al. (2001), Kabelka et al. (2004), Smalley et al. (2004), Zhang et al. (2004), and Wang et al. (2004) identified a yield QTL in the same region on LG C2 where we also mapped a QTL for yield in the Kenwood population. In our study, the yield QTL on LG C2 was also associated with delayed maturity, which was consistent with the results of Specht et al. (2001) and Wang et al. (2004). Specht et al. (2001) noted that the maturity locus E1 is located on this region and they suggested that the yield QTL was this maturity gene. Orf et al. (1999a) and Kabelka et al. (2004) did not observe the association of the yield QTL on LG C2 with delayed maturity. Our results and those of Orf et al. (1999a) and Kabelka et al. (2004) showed that the yield-increasing QTL allele originated from the PI parent, and Specht et al. (2001) mapped the yield QTL using a cross between two PIs. Wang et al. (2003) reported that the yield-increasing QTL allele originated from the domestic parent, and Smalley et al. (2004) noted that the QTL alleles associated with high yield were common to PIs and elite lines.
Using a recombinant inbred line mapping population from a cross between Minsoy x ‘Noir 1’, Orf et al. (1999a) identified a yield QTL in the same region where we detected a QTL for yield on LG J in the Kenwood population. Based on the integrated soybean map (Song et al., 2004), Satt405, where we mapped the QTL in the Kenwood population, is 33 cM from Satt215 and 56 cM from Satt547, where we mapped the yield QTL on LG J in the Beeson 80 population. This indicates that they are independent QTL. Orf et al. (1999a) reported that the yield QTL was associated with plant height, which we did not observe in our study.
The marker Satt313, which was associated with a QTL for yield on LG L in the Kenwood population, also mapped a seed weight QTL in a cross between the cultivars Ma. Belle x Proto (Csanadi et al., 2001). Smalley et al. (2004) reported a yield QTL in the same region on LG L marked by Satt462. Since Satt462 and Satt313 are 6 cM apart on the integrated map (Song et al., 2004), these QTL may not be independent. Csanadi et al. (2001) and Smalley et al. (2004) also detected a seed weight QTL and a yield QTL, respectively, in the same region marked by Satt477 on LG O, where we mapped a yield QTL in the Kenwood population. Csanadi et al. (2001) found that the two QTL alleles for seed weight originated from the higher yielding cultivar, Ma. Belle, while Smalley et al. (2004) reported that some QTL alleles for greater yield associated with Satt462 and Satt477 were common to PIs and the elite parents.
The yield QTL on LG A1 in the Lawrence population was in the same region where a QTL associated with seed weight was mapped by Orf et al. (1999a) in the cross of Noir 1 x Archer. The seed weight QTL identified by Orf et al. (1999a) was marked by Satt449 which is 3 cM away from Satt300 on the integrated map (Song et al., 2004). Among the markers reported by Smalley et al. (2004) that were associated with a yield QTL on LG A1, Satt364 was only 2 cM from Satt300. The proximity of these markers indicates that the seed weight QTL reported by Orf et al. (1999a), and the yield QTL detected by us and by Smalley et al. (2004) may be allelic. The QTL allele for increased seed weight identified in the study of Orf et al. (1999a) originated from Archer, while Smalley et al. (2004) reported that the yield-increasing QTL allele mapped by Satt364 originated from the PI parent.
The yield QTL on LG B2 linked to Satt474 in the Lawrence population maps to the same region as the yield QTL mapped by Orf et al. (1999a), Concibido et al. (2003), and Smalley et al. (2004). Kabelka et al. (2004) reported a yield QTL detected by Satt168 on LG B2, which is 20 cM above Satt474. Orf et al. (1999a) and Smalley et al. (2004) reported yield QTL allele that were marked by Satt066, which is 3 cM from Satt474. Concibido et al. (2003) reported that the yield increasing QTL allele in the same LG B2 region was inherited from a Glycine soja PI. They were able to confirm the QTL in the background of the elite parent they originally used to map the QTL. Further testing of the QTL in different genetic backgrounds showed that it had a positive effect in two of six genetic backgrounds. Although Concibido et al. (2003), like us, found that the yield-increasing QTL allele on LG B2 was inherited from exotic germplasm, the results of Orf et al. (1999a), Smalley et al. (2004), and Concibido et al. (2003) indicate that it is also present in some U.S. cultivars.
Specht et al. (2001) reported a yield QTL marked by Satt281 on LG C2, which is 10 cM from Satt640, a marker that mapped a yield QTL in the Lawrence population. The proximity of these markers suggests that the yield QTL mapped in these populations may be allelic. The yield-increasing QTL allele reported by Specht et al. (2001) was inherited from the PI parent, which was consistent with our study. Satt640 is 72 cM above Satt557 (Song et al., 2004), which is associated with a yield enhancing QTL in the Kenwood population.
Smalley et al. (2004) and Kabelka et al. (2004) reported a QTL for yield associated with Satt358 on LG O, which is 15 cM from Satt445, the marker that mapped a yield QTL in the Lawrence population. Csanadi et al. (2001) also detected an association between seed weight and Satt358. Kabelka et al. (2004) noted that the yield increasing QTL allele on LG O originated from the PI parent. However, our results and those of Smalley et al. (2004) suggest that this QTL allele may also be found in some U.S. cultivars.
For the other agronomic trait QTL, Kabelka et al. (2004) also mapped QTL for maturity close to the region marked by Satt675 on LG N. There are no reports of maturity QTL near the regions marked by Satt556 on LG B2, Satt634 on LG D1b, Satt186 on LG D2, and Satt477 on LG O. Specht et al. (2001) also mapped QTL for lodging in the same region on LG L where we mapped a lodging QTL in the Kenwood population marked by Satt232. Lee et al. (1996) reported a lodging QTL on LG G marked by the RFLP marker A378b, which is 13 cM from Satt191 where we mapped a lodging QTL in the Lawrence population. The other QTL for lodging we identified have not been reported in previous studies. Quantitative trait loci for plant height were mapped in the same regions where we identified plant height QTL on LG C2 (Mian et al., 1998; Orf et al., 1999a; Kabelka et al., 2004; Wang et al., 2004), on LG G (Kabelka et al., 2004), and on LG M (Specht et al., 2001; Wang et al., 2004). There were no previous reports of plant height QTL in the Satt547 region on LG J.
The purpose of analyzing QTL x E interaction is to identify QTL that are stable across environments. Li et al. (2003) described three types of QTL x E interactions: i) the inconsistent detection of the QTL across environments, i.e., the QTL is not detected in all environments; ii) variation in the effects of the QTL across environments, i.e., the QTL is expressed strongly in one environment but weak in another; and iii) a QTL has opposite effects in different environments. Stable QTL that are expressed in broad genetic backgrounds are most useful in marker-assisted selection. In our study, the QTL on LG A1 marked by Satt300 in the Lawrence population had the most consistent effects across environments. Some of the yield-associated QTL showing the greatest QTL x E effects may actually be maturity or disease resistance QTL. For example, the yield QTL marked by Satt547 on LG J had significant QTL x E interaction in 2003, in 2004, and across environments. This QTL is likely a BSR resistance gene locus, and the QTL x E interaction may be the result of differing levels of disease in field environments. In addition, yield QTL marked by Satt557 and Satt477 are associated with maturity and may be maturity QTL that affect yield. It is likely the QTL x E interaction was the result of the maturity differences between the homozygous groups having an inconsistent effect on yield, depending on the specific environmental conditions. In general, the QTL x E interactions detected in our study were largely due to absent/undetectable or weak QTL effects in some environments, and variation in the magnitude of QTL effects. In our study, no QTL for yield showed opposite effects in different environments.
The significance of epistasis in QTL mapping has been underscored by Wang et al. (1999), Liao et al. (2001), and Zhuang et al. (2002). Wang et al. (1999) showed that the precision of QTL mapping is greatly enhanced by including in the QTL model loci exhibiting epistatic interactions. Loci exhibiting favorable epistatic combinations are also targets of marker-assisted selection (Coaker and Francis, 2004). Significant epistatic interactions have been reported for protein and oil content (Lark et al., 1994), and plant height and yield (Lark et al., 1995; Orf et al., 1999b) in soybeans. In our study, a total of nine epistatic interactions out of 3570 digenic combinations was detected in the Kenwood population. The failure to detect epistasis in the Beeson 80 and Lawrence populations may be due to the limited number of markers used in those populations. Liao et al. (2001) described three types of epistasis: i) interactions between two QTL; ii) interactions between a QTL and a "background" locus without an additive effect; and iii) interaction between "complementary" loci or loci exhibiting epistatic effects only. Most of the significant digenic interactions identified in our study were type 3 epistasis (Liao et al., 2001). Satt477 was the only marker significantly associated with yield and was involved in an epistatic interaction in the Kenwood population. The consistent detection of the interaction between Satt197 and Satt257 across years indicates that the regions near these markers, particularly that of Satt257, were important in yield expression in the Kenwood population through epistatic interactions. However, Satt257 was also associated with a QTL for greater lodging, which may limit its potential as a candidate for a marker-assisted selection program for yield improvement. Orf et al. (1999b) reported that the region marked by the RFLP marker B172_2 on LG A2 was involved in a significant epistatic interaction for yield. This marker mapped 4 cM from Satt187 (Song et al., 2004), which was shown to interact epistatically with Satt257 in 2003 and in the means across environments. The proximity of these markers indicates that they are detecting the same region involved in epistatic interaction for yield. Wang et al. (2004) identified a region on LG D1b marked by Satt189 that was involved in an epistatic interaction for yield in one of the BC populations they studied. This marker mapped 22 cM from Satt703, which showed significant interaction in 2003. Further studies are needed to verify whether the region on LG D1b identified in our study and the study of Wang et al. (2004) are the same.
The thirteen QTL for yield we identified were on the same or close to the region where yield QTL have been reported. Although we determined that the yield-increasing allele for eight yield QTL originated from the PI parents, the majority were also detected in U.S. cultivars in previous studies. Our results support the findings that the current commercial gene pool is more diverse than would have been predicted by the number of contributing ancestors (Brown-Guedira et al., 2000; Li et al., 2001), and that identifying new yield enhancing alleles from soybean exotic germplasm may be more difficult than anticipated. Wang et al. (2004) also concluded that it is difficult to identify new useful genetic diversity from G. soja. Additional studies involving different PIs will be important to expand the search for unique yield enhancing alleles that are not present in the U.S. soybean cultivars. The yield QTL detected, particularly those with large effects, should be confirmed, to justify using them in a marker-assisted selection program. Confirmation in different genetic backgrounds could be implemented by selecting lines homozygous for the positive alleles at the regions of interest and crossing them with cultivars or unrelated lines. Results of our study also support the hypotheses that QTL x E interaction and epistasis influence soybean yield. Interacting loci that exhibit significant positive effect could also be used in a marker-assisted selection program to improve a trait of interest, such as yield.
Received for publication January 2, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||
