Crop Science Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 24 February 2006
Published in Crop Sci 46:721-734 (2006)
© 2006 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (5)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Tang, S.
Right arrow Articles by Knapp, S. J.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Tang, S.
Right arrow Articles by Knapp, S. J.
Agricola
Right arrow Articles by Tang, S.
Right arrow Articles by Knapp, S. J.
Related Collections
Right arrow Crop Genetics
Right arrow Sunflower

GENOMICS, MOLECULAR GENETICS & BIOTECHNOLOGY

Quantitative Trait Loci for Genetically Correlated Seed Traits are Tightly Linked to Branching and Pericarp Pigment Loci in Sunflower

Shunxue Tanga, Alberto Leonb, William C. Bridgesc and Steven J. Knappa,*

a Center for Applied Genetic Technologies, 111 Riverbend Road, The University of Georgia, Athens, GA, 30602, USA
b Advanta Seeds, Balcarce Research Station, Ruta 226, KM 60.3 (7620), Balcarce PCIA DE BS. AS., ARGENTINA
c Department of Experimental Statistics, Clemson University, Clemson, SC 29634

* Corresponding author (sjknapp{at}uga.edu)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The seed oil concentrations of large-seeded, low-oil and small-seeded, high-oil sunflower (Helianthus annuus L.; x = 17) cultivars differ by 180 to 280 g kg–1. We identified quantitative trait loci (QTL) for seed oil and other seed traits in a low- x high-oil (RHA280 x RHA801) recombinant inbred line (RIL) mapping population segregating for apical branching (B), phytomelanin pigment (P), and hypodermal pigment (Hyp) loci. B, Hyp, and P mapped to linkage groups 10, 16, and 17, respectively. The seed oil concentrations of RHA280 and RHA801 were 254 and 481 g kg–1, respectively. Composite interval mapping (CIM) identified 40 QTL for seed oil concentration, 100-seed weight, seed length, width and depth, kernel and pericarp weight, and kernel-to-pericarp weight ratio in 14 DNA marker intervals on 10 of 17 linkage groups. Twenty-four of the QTL were tightly linked to B, P, and Hyp and may have been partly or wholly caused by the pleiotropic effects of B, P, and Hyp. Multilocus QTL analyses were performed using B, P, Hyp, and four DNA marker loci as independent variables in mixed linear models. Seventy percent of the additive effects (39/56) and 42% of the additive x additive and additive x additive x additive effects (189/448) were significant (p < 0.05). The linked, pleiotropically acting, and epistatically interacting QTL identified for seed traits in RHA280 x RHA801 were presumably targeted by selection in the transition from large-seeded, low-oil to small-seeded, high-oil cultivars in sunflower.

Abbreviations: INDEL, insertion-deletion • LG, linkage group • QTL, quantitative trait locus • RIL, recombinant inbred line • SSR, simple sequence repeat


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
THE ACHENES (seeds) of cultivated or common sunflower (Helianthus annuus L.) are a rich source of oil and an historically important source of food and pigment for Native Americans (Heiser, 1951, 1977). The seeds of unbranched (single-headed) Native American land races, the earliest cultigens, are significantly larger than the seeds of wild populations, partly because unbranched biotypes have larger capitula and seeds than branched biotypes, and partly because sunflower was domesticated by selecting for larger seeds (Ross, 1939; Heiser, 1951, 1977; Dedio, 1980). Shellable, large-seeded, low-oil (confectionery) land races developed in eastern Europe were introduced to North America in 1880 and had seed oil concentrations in the 260 to 280 g kg–1 range. Mennonite, an early confectionery cultivar, CM 612 (PI 546351), an inbred directly isolated from Mennonite (Dedio and Rashid, 1991), and numerous other confectionery cultivars and inbred lines have seed oil concentrations in the 240 to 290 g kg–1 range (http://www.ars-grin.gov; verified 2 December 2005).

Unshellable, small-seeded, high-oil (oilseed) cultivars were developed between 1920 and 1955 by direct selection for increased seed oil concentration and indirect selection for small seeds and thin pericarps (hulls) in open-pollinated populations (Putt, 1940, 1997; Heiser, 1951, 1977; Pustovoit, 1964). Cultivars producing 380 g kg–1 of seed oil may have been developed as early as 1915; however, the seed oil concentrations of cultivars commonly grown between 1900 and 1915 reportedly ranged from 200 to 300 g kg–1 (Putt, 1997). Selection in open-pollinated populations increased seed oil concentrations to 430 g kg–1 by 1935 and 490 g kg–1 by 1955 (Pustovoit, 1964; Heiser, 1977; Seiler, 1985, 1994; Putt, 1997; Seiler and Brothers, 1999), thereby greatly increasing the economic importance of sunflower as an oilseed and further differentiating the confectionery and oilseed market classes (Pustovoit, 1964; Putt, 1997).

Seeds of confectionery and oilseed cultivars are distinguished by differences in shellability, hull color, seed weight and morphology, and kernel-to-pericarp weight ratio, in addition to seed oil concentration (Beard, 1981; Seiler, 1997). The seeds of confectionery cultivars are typically gray or white, black or brown striped, and easily dehulled or shelled, whereas the seed of oilseed cultivars are typically black (dark purple) and difficult to dehull. Seeds of confectionery cultivars are larger and have lower kernel-to-pericarp weight ratios than seeds of oilseed cultivars. Hull color differences are produced by pigments in the epidermal, hypodermal, and phytomelanin layers (Johnson and Beard, 1977; Leon et al., 1996; Miller and Fick, 1997). The epidermis is either unpigmented, solid brown or black, or black- or brown-striped, the hypodermis is either anthocyanin pigmented (black) or unpigmented, and phytomelanin is either present (black) or absent (Johnson and Beard, 1977; Leon et al., 1996; Miller and Fick, 1997). Several hull pigment loci have been identified through phenotypic analyses of mutations (reviewed by Miller and Fick, 1997). The allelism of many of the mutations is not known, and only Hyp, a hypodermis pigment locus, has been genetically mapped (Leon et al., 1996).

The first high-oil cultivars were apparently developed by introgressing allelic diversity from wild populations into open-pollinated, low-oil cultivars (Pustovoit, 1964; Heiser et al., 1969; Semelczi-Kovacs, 1975; Putt, 1997). Selection within domesticated, low-oil populations per se cannot be completely ruled out—land race selec tion between 1880 and 1915 could have played a role in advancing sunflower as an oilseed (Putt, 1997). Significant genetic variability for seed oil concentration is present in wild sunflower germplasm, e.g., seed oil concentrations ranged from 191 to 355 g kg–1 among 340 wild H. annuus populations screened by the USDA (http://www.ars-grin.gov). The upper end of the range was even greater among 93 wild populations of H. anomalus S. F. Blake, H. petiolaris Nutt., H. debilis Nutt., and other taxa—minimum, mean, and maximum seed oil concentrations were 262, 346, and 457 g kg–1, respectively.

Once hybrid seed production systems were developed (Leclercq, 1969; Kinman, 1970), the focus in sunflower breeding rapidly shifted to the development of high-oil inbred lines from an assortment of open-pollinated, high-oil populations and cultivars, the progenitors of most of the early high-oil inbred lines on which the hybrid seed industry was initially built (Korell et al., 1992; Cheres and Knapp, 1998). While present-day hybrids commonly produce 440 to 490 g kg–1 of seed oil, CM 612 (PI 546351), CM 630 (PI 566828), and other inbred lines producing 500 to 530 g kg–1 of seed oil have been described (Dedio and Rashid, 1991, 1994; http://www.ars-grin.gov). Our goal was to identify phenotypic and quantitative trait loci (QTL) underlying genetic variability for pericarp pigments, seed oil concentration, and other seed traits in a recombinant inbred line (RIL) mapping population (Tang et al., 2002) developed from a hybrid between large-seeded, low-oil (RHA280) and small-seeded, high-oil (RHA801) inbred lines developed in the early period of single-cross hybrid breeding in sunflower (Fick et al., 1974a; Roath et al., 1981).


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Genetic Mapping and Phenotyping
Genetic analyses were performed on 173 F7 recombinant inbred lines (RILs) developed by single-seed-descent from a cross between RHA280 (an unbranched, confectionery, fertility restorer line) and RHA801 (a branched, oilseed, fertility restorer line) (Tang et al., 2002). RHA280 has large, brown-striped seeds and low seed oil concentrations (Fick et al., 1974a), whereas RHA801 has small, black seeds and high seed oil concentrations (Roath et al., 1981). The RILs were presumed to be segregating for an apical branching gene (B) described by Putt (1964) and the B locus genotypes for RHA280 and RHA801 were presumed to be BB and bb, respectively.

The RILs were field grown at Corvallis, OR, in the summers of 2000 and 2001 in 6.1-m-long rows spaced 0.9 m apart. Two replications of the RILs and parent inbred lines were planted in randomized complete blocks in May of both years. Fifty-six kilograms per hectare of 20N-10P-15K- 7S fertilizer was broadcast preplant and incorporated during the primary tillage operation. The seed bed was prepared by disking and harrowing, and seeds were planted 2.54 cm deep. The within-row seeding rate was 40 seeds per 6.1 m. Plants were hand-thinned to produce a within row spacing of one plant per 0.3 m 2 wk postemergence. Capture [bifenthrin = (2-methyl-3-phenyl-phenyl)methyl 3-(2-chloro-3,3,3-trifluoro-prop-1-enyl)-2,2-dimethyl-cyclopropane-1-carboxylate, 0.11 kg ai ha–1] and Hoelon {diclofop-methyl = methyl (RS)-2-[4-(2,4-dichlorophenoxy)phenoxy]propionate, 2.35 L ha–1} were applied postplant, preemergence mixture for broadleaf and grass weed control and were activated by a minimum of 1.27 to 2.54 cm of water applied through sprinkler irrigation, rainfall, or both. Subsequent to thinning, 100.8 kg per ha of N was applied by banding urea (46N-0-0). The crop was sprinkler irrigated at 2-wk intervals throughout the first 12 wk of the growing season; 1.28 to 2.54 cm of water was applied at each interval. Capitula were harvested in late September of both years, dried in a forced-air gas seed drier for 48 to 72 h at 42 to 44°C, threshed, and cleaned.

The RILs were phenotyped at the onset of flowering for presence and absence of apical branching (20 plants per RIL per replication were phenotyped). Four to 12 physiologically mature seeds per RIL were visually phenotyped for hypodermal and phytomelanin pigments in the pericarp by dissecting through successive pericarp cell layers (epidermis, hypodermis, and phytomelanin) as described by Leon et al. (1996). Capitula were harvested from 10 individuals per RIL per replication. Primary capitula were harvested from unbranched RILs, whereas primary and two secondary capitula were harvested from branched RILs. Kernel weight (kwt), pericarp weight (pwt), and kernel-to-pericarp weight ratio (kpr = kwt/pwt) were measured on 10 seeds per RIL per replication produced in 2000. The 10 seeds per RIL were manually dehulled with a scalpel and separated into kernel and hull fractions. Five-gram samples of seeds were randomly drawn from each replication and pooled into a single 10-g sample for seed oil concentration (soc) analyses by nuclear magnetic resonance (Leon et al., 1995). We measured seed length (sl), width (sw), and depth (sd) on 10 randomly sampled seeds per RIL per replication and 100-seed weight (swt) on 100 randomly sampled seeds per RIL per replication.

We constructed a genetic linkage map for the QTL analysis using three phenotypic loci (B, P, and Hyp) and 203 simple sequence repeat (SSR) or insertion-deletion (INDEL) marker loci selected from the original 577-locus RHA280 x RHA801 genetic linkage map described by Tang et al. (2002) and Yu et al. (2003). The original map was constructed by genotyping 94 RILs; hence, 203 SSR or INDEL markers were genotyped on 79 additional RILs for the present study. The terminal-most or near terminal-most SSR or INDEL marker loci from each linkage group were selected for genotyping in addition to SSR or INDEL marker loci spaced as evenly as possible throughout each linkage group. The genotyping methods and statistical analyses used to construct the map have been described elsewhere (Tang et al., 2002).

Statistical Analyses
The phenotypic distributions for seed traits were tested for normality by the Kolmogorov- Smirnov (D) statistic. D was estimated by PROC UNIVARIATE of the Statistical Analysis System (SAS) (http://www.sas.com). Progeny-mean (RIL-mean) heritabilities (h2) and additive genetic correlations (rG) were estimated for RILs among years and replications for 100-seed weight and seed length, width, and depth, among years for seed oil concentration, and among replications for kernel and pericarp weight and kernel-to-pericarp weight ratio. RILs, years, and replications were identified as random effects for variance and covariance component analyses. Variance components were estimated by the REML method of SAS PROC VARCOMP and covariance components were estimated by the REML method of SAS PROC MIXED as described by Holland et al. (2001). RIL-mean heritabilities (h2) and genetic correlations (rG) were estimated as described by Falconer and MacKay (1996) and Bernardo (2002).

Composite interval mapping (CIM), as implemented in QTL-Cartographer (Zeng, 1993, 1994; Zeng et al., 1999; Basten et al., 2001), was used to scan for additive intralocus QTL effects. CIM analyses were performed on least square means for RILs estimated by SAS PROC MIXED, where RIL effects were fixed and replication, year, and RIL x year interaction effects were random. Tests of significance of RIL, replication, year, and RIL x year effects were performed with Type III F statistics estimated by SAS PROC GLM. Rank correlations between least square means for RILs between years were estimated by SAS PROC CORR. CIM analyses were performed on RIL least square means across years (RIL x year effects were nonsignificant for every trait).

CIM analyses were performed with a 5-cM window width and significant cofactors identified by QTL-Cartographer; 10 significant cofactors were identified and selected for CIM analyses of soc and five significant cofactors were identified and selected for CIM analyses of other seed traits. B, P, and Hyp were used as cofactors for every trait. Likelihood-odds (LOD) thresholds for declaring statistical significance (the presence of QTL) were calculated by permutation testing with 1000 permutations (Churchill and Doerge, 1994). Permutation thresholds ranged from 2.9 to 3.2 for the eight traits. One-LOD support intervals were calculated as described by Conneally et al. (1985) and Lynch and Walsh (1998). The additive effects (a) and phenotypic coefficients of determination (R2P) for individual QTL were estimated by CIM.

Multilocus QTL analyses were performed with seven genetic marker loci as independent variables in mixed linear models (Littel et al., 1991, 1996). The loci selected as independent variables—B, P, Hyp, and four SSR marker loci (ORS371, ORS188, ORS1068, and ORS484)—were tightly linked to QTL identified by CIM for multiple seed traits. Statistical analyses were performed by SAS PROC MIXED, where genotype (G) effects were fixed and RIL nested in genotype (RIL:G), year (Y), replication (R), and genotype x year (G x Y) effects were random. Variance components were estimated by the REML method, Type III sums of squares and F statistics were estimated for genotype effects, least square means were estimated for genotypes, and additive (a), additive x additive (a x a), and additive x additive x additive (a x a x a) effects were estimated by ESTIMATE statements.

Statistical analyses were performed on homozygotes only because heterozygotes were rare and produced severely unbalanced data and missing cells. We performed an incomplete 27 factorial analysis by estimating first-, second-, and third-order effects (a, a x a, and a x a x a, respectively) among the seven genetic marker loci. RIL:G and G x Y variance components were nonsignificant and were subsequently pooled with other random residual effects to maximize error degrees of freedom (dfe); dfe were 215 for seed oil concentration, 540 for 100-seed weight and seed length, width, and depth, and 237 for kernel and pericarp weight and kernel-to-pericarp weight ratio. Genetic coefficients of determination were estimated for a effects only (R2G = SSG/SSRIL) and a, a x a, and a x a x a effects combined (R2G* = SSG*/SSRIL), where SSG is the sum of squares for a effects, SSG* is the sum of squares for a, a x a, and a x a x a effects, SSRIL is the sum of squares among RILs, and SSG, SSG*, and SSRIL were estimated by SAS PROC GLM (Littel et al., 1996).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Genetic Mapping
We developed a 203-locus genetic map for the QTL analyses presented herein (Fig. 1 ) using 173 RHA280 x RHA801 RILs and SSR and INDEL marker loci selected from a 577-locus genetic map previously constructed using 94 of the 173 RHA280 x RHA801 RILs (Tang et al., 2002; Yu et al., 2003). The terminal-most or near terminal-most genetic marker loci on each of the 17 linkage groups were genotyped in addition to genetic marker loci spaced as evenly as possible throughout each of the 17 original linkage groups. Whenever possible, the highest quality single-locus SSR and INDEL marker loci identified in previous analyses (Tang et al., 2002, 2003; Tang and Knapp, 2003; Yu et al., 2003) were selected for genotyping. Groups and locus orders for SSR and INDEL marker loci were identical between the two genetic maps. The 10 linkage groups where QTL were found are shown in Fig. 1.


Figure 1
Figure 1
View larger version (39K):
[in this window]
[in a new window]
  		Fig. 1. Quantitative trait loci (vertical bars are 1.0-LOD support intervals) identified by composite interval mapping in the RHA280 x RHA801 recombinant inbred line mapping population for seed oil concentration (soc), 100-seed weight (swt), seed length (sl), width (sw), and depth (sd), kernel weight (kwt), pericarp weight (pwt), and kernel-to-pericarp weight ratio (kpr).

 
RHA280 and RHA801 differed and the RILs segregated for several quantitative seed traits, apical branching, and hull hypodermal and phytomelanin pigments but not for hull epidermal pigments (dark brown stripes were present in the epidermis in both inbred lines) (Tables 1 and 2 and Fig. 2 and 3) . The hull pigment genotypes of the parents and hull pigment loci segregating among RHA280 x RHA801 RILs were not known a priori but were inferred from hypodermal and phytomelanin pigment segregation ratios (see below). The RILs segregated for an apical branching locus (B; Putt, 1964), a hull hypodermis pigment locus (Hyp; Leon et al., 1996), and a hull phytomelanin pigment locus (P) (Fig. 2). The confectionery inbred line (RHA280) had an opaque white hypodermis (Hyp Hyp) and unpigmented phytomelanin (p p), whereas the oilseed inbred line (RHA801) had a transparent hypodermis (hyp hyp) and black phytomelanin (P P).


View this table:
[in this window]
[in a new window]
  		Table 1. Hull pigment, branching, and seed phenotypes for the confectionery (RHA280) and oilseed (RHA801) inbred line parents of the RHA280 x RHA801 recombinant inbred line mapping population.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Least square means, minimums, and maximums for 173 recombinant inbred lines (RILs), RIL-mean heritabilities (h2), and genetic coefficients of determination for additive effects (R2G) or additive, additive x additive, and additive x additive x additive effects (R2G*) of genetic marker loci (B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484) for seed traits segregating in the RHA280 x RHA801 RIL mapping population.

 

Figure 2
View larger version (54K):
[in this window]
[in a new window]
  		Fig. 2. Seeds for parents of the RHA280 x RHA801 recombinant inbred line (RIL) mapping population (labeled 280 and 801, resepectively) and RHA280 x RHA801 RILs from the lower and upper ends of the 100-seed weight distributions within each B x P x Hyp genotypic class (upper and lower panels, respectively), where B, P, and Hyp are apical branching, phytomelanin pigment, and hypodermal pigment loci, respectively (RIL seeds are identified by numbers in the bottom left-hand corner of each photograph).

 

Figure 3
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 3. Least square means for RHA280 and RHA801 (solid triangle) and branched (open circle) and unbranched (open inverted triangle) RHA280 x RHA801 recombinant inbred lines segregating for seed oil concentration, kernel-to-pericarp weight ratio, and 100-seed weight (rG = additive genetic correlation).

 
The observed segregation ratio for B, 85 branched (b b): 2 segregating (B b): 86 unbranched (B B), was not significantly different from the expected segregation ratio ({chi}2 = 0.01, p = 0.91) for F6:7 RILs (85.2 b b: 2.7 B b: 85.2 B B). The B locus mapped to LG 10 (Fig. 1). The observed segregation ratio for Hyp, 83 opaque white (Hyp Hyp): 2 segregating (Hyp hyp): 88 transparent (hyp hyp), was not significantly different from the expected segregation ratio ({chi}2 = 0.41; p = 0.65). The Hyp locus mapped to LG 16 (Fig. 1). The observed segregation ratio for P, 71 non- pigmented (p p): 1 segregating (P p): 101 black (P P), was significantly different from the expected segregation ratio ({chi}2 = 6.9, p = 0.008) because of an excess of P P RILs. The segregation ratios for SSR marker loci tightly linked to P had similarly distorted segregation ratios. The P locus mapped to LG 17 (Fig. 1).

Heritabilities and Genetic Correlations
The phenotypic distributions were normal for soc, sl, sw, and sd and approximately normal but slightly positively skewed for swt, kwt, pwt, and kpr (the phenotypic distributions for soc x kpr and soc x swt are shown in Fig. 3 for branched and unbranched RILs). Significant genetic variability was observed among RILs for every trait. Year effects were significant for every trait, whereas RIL x year interaction effects were nonsignificant for every trait measured in both years; kwt, pwt, and kpr were only measured 1 yr. The rank correlations for RIL least square means between years were significant, high, and positive for the five traits measured in both years (0.85 for soc, 0.88 for swt, 0.92 for sl, 0.90 for sw, and 0.85 for sd). RIL-mean heritabilities were exceptionally high and ranged from 0.92 to 0.98 (Table 2).

Strong genetic correlations were observed among the eight seed traits (Table 3 and Fig. 3). Oil concentration was negatively genetically correlated with pericarp weight (–0.75) and 100-seed weight (–0.58) and positively genetically correlated with kernel-to-pericarp weight ratio (0.85). Kernel and pericarp weight were positively genetically correlated with seed length, width, and depth (0.76–0.88) and 100-seed weight (0.94).


View this table:
[in this window]
[in a new window]
  		Table 3. Genetic correlations among seed traits segregating in the RHA280 x RHA801 recombinant inbred line mapping population.

 
Positive transgressive segregates were not observed for soc, kpr, or swt and none of the RILs produced more oil than the high-oil parent (RHA801). The RIL from the upper end of the soc x kpr and soc x swt distributions (RIL263) was homozygous for RHA801 alleles among the seven loci tested in the multilocus QTL model (Table 1; Fig. 2 and 3). Gaps were observed at the upper ends of the soc x kpr and soc x swt distributions between the high oil parent (RHA801) and the closest RILs, especially for kernel-to-pericarp weight ratio (Fig. 3). These gaps illustrate the difficulty of recovering progeny with phenotypes equal or transgressive to the small-seeded, high-oil parent (e.g., RHA801) in crosses to large-seeded, low-oil inbreds (e.g., RHA280). The loss of one or two favorable QTL alleles produced a significant drop-off in both kpr and soc. Conversely, negative transgressive segregates were observed for kpr and soc and RIL means for kpr were skewed towards the mean of the low oil parent (RHA280) (Fig. 3).

QTL Identified by Composite Interval Mapping
Using CIM, we identified 40 QTL for eight seed traits in 14 marker intervals on 10 linkage groups (Fig. 1 and Table 4). Three-fourths of the QTL were clustered on linkage groups (LG) 5 (five QTL), 10 (eight QTL), 16 (six QTL), and 17 (10 QTL). QTL on seven linkage groups affected multiple traits (Fig. 1). The QTL identified on LG 10, 16, and 17 were centered on or near B, Hyp, and P, respectively, whereas QTL identified on LG 1, 4, 5, and 9 were centered on or near four SSR marker loci (ORS371, ORS1068, ORS484, and ORS188, respectively). Single-trait QTL were identified for sl on LG 8, sd on LG 12, and swt on LG 14 (Fig. 1). Four to six QTL were identified for each trait and collectively explained 52.3 to 77.5% of the phenotypic variability per trait (individual QTL explained 3.0 to 52.5% of the phenotypic variability per trait) (Table 4).


View this table:
[in this window]
[in a new window]
  		Table 4. Linkage groups (LG), likelihood odds (LOD) scores and positions for quantitative trait locus (QTL) peaks on the genetic map (cM), 1.0-LOD support intervals, genetic marker intervals spanning QTL peaks, closest genetic marker locus to the QTL peak, phenotypic coefficients of determination (R2P), and additive genetic effects (a) for seed oil concentration (soc; g kg–1), 100- seed weight (swt; g), seed length (sl; mm), width (sw; mm), and depth (sd; mm), 10-kernel weight (kwt; g), 10-pericarp weight (pwt; g), and kernel-to-pericarp weight ratio (kpr) QTL segregating in the RHA280 x RHA801 recombinant inbred line mapping population, where a = (yRHA801 – yRHA280)/2, yRHA280 is the RHA280 homozygote mean, and yRHA801 is the RHA801 homozygote mean.

 
B-linked QTL had the largest effects on every trait except kernel-to-pericarp ratio (Table 4). The effects of P and Hyp-linked QTL were less than the effects of B-linked QTL for every trait other than kernel-to-pericarp ratio, where the effects of the three loci were nearly equal. Branching (b b) increased seed oil concentration and kernel-to-pericarp weight ratio and decreased 100-seed weight and seed length, width, and depth (Fig. 3). Seeds of RILs with an opaque hypodermal layer (Hyp Hyp) and no phytomelanin layer (p p) were larger, had lower kernel-to-pericarp weight ratios, and produced less seed oil than seeds of RILs with a transparent hypodermal layer (hyp hyp) and black phytomelanin layer (P P) (Fig. 2).

Six QTL were identified on LG 1, 4, 9, 10, 16, and 17 for seed oil concentration (Fig. 1 and Table 4). The QTL individually explained 3.1 to 22.5% and collectively explained 55.7% of the phenotypic variability for soc. QTL for soc on LG 10, 16, and 17 were centered on the B, Hyp, and P loci, respectively, and QTL alleles for increased oil were transmitted by the oilseed parent (RHA801). B-, Hyp-, and P-linked QTL (soc10.1, soc16.1, and soc17.1) explained 22.5, 10.5, and 3.1% of the phenotypic variability, respectively. R2P for the third largest QTL (soc9.1) was 9.1%.

Five QTL were identified on LG 5, 9, 10, 14, and 17 for 100-seed weight (Fig. 1 and Table 4). The QTL collectively explained 73.4% of the phenotypic variability. B- and P-linked QTL (swt10.1 and swt17.1) explained 52.5 and 7.4% of the phenotypic variability for swt, respectively. The minor QTL swt9.1 and swt14.1 explained 3.2 and 3.6% of the phenotypic variability, respectively. The RHA280 allele from four of the five QTL increased 100-seed weight, whereas the RHA280 allele for one of the five QTL (swt5.1) decreased 100-seed weight. The swt5.1 QTL had the third largest effect.

Four QTL were identified for kwt on three linkage groups (LG 5, 10, and 16) and collectively explained 57.3% of the phenotypic variability (Fig. 1 and Table 4). Two linked QTL were identified on LG 16 (kwt16.1 and kwt16.2). Favorable alleles for the B-linked QTL (kwt10.1) were transmitted by the unbranched, low-oil parent (RHA280), whereas favorable alleles for the other three QTL (kwt5.1, kwt16.1 and kwt16.2) were transmitted by the branched, high-oil parent (RHA801). Similarly, five QTL were identified for pwt on four linkage groups (LG 4, 5, 10, and 17) and collectively explained 77.5% of the phenotypic variability. Favorable alleles were transmitted by the low-oil parent for one of the five QTL (pwt5.1) (Table 4). Two linked QTL were identified for pwt on LG 17 (pwt17.1 and pwt17.2). Coincident QTL were identified for swt, kwt, and pwt in DNA marker intervals on linkage groups 5 (kwt5.1, pwt5.1, and swt5.1), 10, (kwt10.1, pwt10.1, and swt10.1), and 17 (pwt17.2 and swt17.1) (Fig. 1). By contrast, kwt and pwt QTL were not identified in two DNA marker intervals where small-effect swt QTL were found (swt9.1 and swt14.1) (Table 4).

Five QTL were identified for kernel-to-pericarp weight ratio on LG 1, 4, 10, 16, and 17 and collectively explained 52.3% of the phenotypic variability. RHA801 alleles for every QTL increased kpr (Fig. 1 and Table 4). The five kpr QTL were coincident with five of six SOC QTL and, as noted earlier, the effects of the B-, Hyp-, and P-linked QTL (kpr10.1, kpr16.1, and kpr17.1) were virtually equal and individually explained 13.5 to 15.2% of the phenotypic variance for kpr.

Five, 4, and 6 QTL were identified for sl, sw, and sd, respectively, and collectively explained 75.9, 73.9, and 73.1% of the phenotypic variance, respectively (Fig. 1 and Table 4). QTL for sl, sw, and sd were clustered on LG 5, 10, 16, and 17 and overlapped SOC, swt, kwt, pwt, and kpr QTL. QTL for sl, sw, and sd on LG 5 and 16 had positive additive effects (alleles from the high- oil parent increased seed length, width, and depth), as was observed for overlapping swt, kwt, and pwt QTL on LG 5 and 16 where alleles from the high-oil parent increased 100-seed, kernel, and pericarp weight. The three B-linked QTL for sl, sw, or sd (sl10.1, sw10.1, and sd10.1) explained 41.7, 44.0, and 43.6% of the phenotypic variability and the five P-linked QTL for sl, sw, or sd (sl17.1, sl17.2, sw17.1, sw17.2, and sd17.1) explained 10.5, 9.2, 13.7, 12.7, and 13.4% of the phenotypic variability.

Multilocus QTL Analyses
Multilocus QTL analyses were performed with phenotypic and DNA marker loci (B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484) identified by CIM to be tightly linked to QTL for multiple seed traits (Fig. 1). The array of marker genotypes observed among the RILs was insufficient for performing a complete 27 factorial analysis but sufficient for performing an incomplete 27 factorial analysis with first-, second-, and third-order (a, a x a, and a x a x a, respectively) effects among B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484 (Fig. 4 Go –6 )–every a, a x a, and a x a x a effect was estimable, 89 out of 128 possible homozygous genotypes were observed among 153 RILs, 20 RILs were heterozygous for one or more loci, and less than half of the genotypes were replicated (were observed in two or more RILs). Using a significance threshold of p ≤ 0.05, 39 out of 56 a, 70 out of 168 a x a, and 119 out of 280 a x a x a effects were significant among the eight seed traits (Fig. 4Go–6). Using a more stringent significance threshold (p ≤ 0.0001), 24 a, 11 a x a, and 36 a x a x a effects were significant. R2G* ranged from 0.63 to 0.82; hence, two-thirds to fourth-fifths of the phenotypic variability among RILs was associated with QTL (Table 2). B-, P-, and Hyp-linked QTL alone were associated with more than half of the phenotypic variability among RILs–R2G* ranged from 0.50 to 0.63 for the eight seed traits in complete 23 factorial analyses performed with only B, P, and Hyp as the independent variables.


Figure 4
View larger version (35K):
[in this window]
[in a new window]
  		Fig. 4. Significant additive, additive x additive, and additive x additive x additive effects among genetic marker loci (B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484) for seed oil concentration and 100-seed weight in the RHA280 x RHA801 recombinant inbred line mapping population, where a = (yRHA801 – yRHA280)/2 is the additive effect, yRHA801 is the least square mean for genotypes homozygous for the RHA801 allele, yRHA280 is the least square mean for genotypes homozygous for the RHA280 allele, and the significance threshold for Type III F tests was p ≤ 0.05.

 

Figure 5
View larger version (36K):
[in this window]
[in a new window]
  		Fig. 5. Significant additive, additive x additive, and additive x additive x additive effects among genetic marker loci (B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484) for seed length, width, and depth in the RHA280 x RHA801 recombinant inbred line mapping population, where a = (yRHA801 – yRHA280)/2 is the additive effect, yRHA801 is the least square mean for genotypes homozygous for the RHA801 allele, yRHA280 is the least square mean for genotypes homozygous for the RHA280 allele, and the significance threshold for Type III F tests was p ≤ 0.05.

 

Figure 6
View larger version (35K):
[in this window]
[in a new window]
  		Fig. 6. Significant additive, additive x additive, and additive x additive x additive effects among genetic marker loci (B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484) for kernel weight, pericarp weight, and kernel-to-pericarp weight ratio in the RHA280 x RHA801 recombinant inbred line mapping population, where a = (yRHA801 – yRHA280)/2 is the additive effect, yRHA801 is the least square mean for genotypes homozygous for the RHA801 allele, yRHA280 is the least square mean for genotypes homozygous for the RHA280 allele, and the significance threshold for Type III F tests was p ≤ 0.05.

 
The proportion of the phenotypic variability among RILs associated with intergenic interactions among the QTL was estimated by 1–R2G/R2G* and ranged from 0.21 to 0.33 (Table 2). The number of significant a x a and a x a x a effects (p < 0.05) was lowest for kernel-to- pericarp weight ratio (six), next lowest for seed oil concentration (nine), and greatest for 100-seed weight (33) (Fig. 4Go–6). While positive and negative epistatic interactions were observed for every trait, most were negative (intragenic and intergenic interaction effects had opposite signs) and crossover in nature. The sign differences we observed between intragenic (a) and intergenic (a x a and a x a x a) effects (Fig. 4Go–6) and between sums of significant (p < 0.05) intragenic and intergenic effects (Table 5) illustrate the predominantly negative or antagonistic (Holland, 2001) nature of epistasis among QTL for seed traits in RHA280 x RHA801. If the intragenic effect of a QTL was positive (increased the mean), most of the intergenic effects between the QTL and other QTL were negative (decreased the mean). Sums of significant intergenic QTL effects were greater than sums of significant intragenic QTL effects for six of the eight traits (swt, sl, sw, sd, kwt, and pwt) (Table 5). Conversely, sums of significant intergenic QTL effects were less than sums of significant intragenic QTL effects for seed oil concentration and kernel-to-pericarp weight ratio. The intergenic to intragenic effect ratio was lowest for seed oil concentration (0.03) and greatest for kernel weight (7.07) and seed length (13.85) (Table 5).


View this table:
[in this window]
[in a new window]
  		Table 5. Sum of statistically significant (p < 0.05) additive (intragenic) and additive x additive and additive x additive x additive (intergenic) effects for seven genetic marker loci (B, P, Hyp, ORS371, ORS188, ORS1068, and ORS484) on seed traits segregating in the RHA280 x RHA801 recombinant inbred line mapping population.

 
Some of the QTL identified by CIM for one or more seed traits, when incorporated into multilocus models with intra- and intergenic effects, were found to affect other seed traits through intergenic interactions and were discovered to have a more important role in producing genetic variability than was found by using models with intragenic effects alone (Table 4 and Fig. 4Go–6). Hyp- and ORS371-linked QTL for seed weight, length, width, and depth affected other traits through predominantly negative epistatic interactions (Fig. 4Go–6). ORS484 was selected for multilocus QTL analyses because of tight linkage to QTL for 100-seed weight, seed length and depth, and kernel and pericarp weight (Fig. 1 and Table 4). Contrary to other QTL, ORS484 alleles transmitted by the high-oil parent increased 100-seed weight, seed length, width, and depth, and pericarp weight (Fig. 1 and 4Go–6). The additive effect of ORS484 on seed oil concentration was nonsignificant because of the canceling effect of the crossover additive x additive interaction effect; however, ORS484 affected seed oil concentration through one positive (ORS371 x ORS188 x ORS484) and two negative (B x ORS484 and ORS371 x ORS484) epistatic interactions (Fig. 4).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Selection for larger seeds was integral to sunflower domestication (Heiser, 1951, 1977; Heiser et al., 1969; Burke et al., 2002); however, direct selection for increased seed oil in early oilseed sunflower breeding programs indirectly selected for smaller seeds and shifted the phenotype toward the wild type (Heiser, 1951, 1977; Pustovoit, 1964; Heiser et al., 1969; Semelczi-Kovacs, 1975; Miller and Fick, 1997; Putt, 1997; Seiler, 1997). Our analyses identified several pleiotropically acting and epistatically interacting QTL, in addition to linked QTL, for seed traits targeted by selection in the transition from large-seeded, low-oil to small-seeded, high-oil cultivars (Fig. 1GoGoGoGo–6). Selection for increased seed oil concentration seems to have operated on multiple QTL affecting several genetically correlated seed traits. Thus far, 54 QTL have been identified for eight seed traits on 14 linkage groups (20 DNA marker intervals) in RHA280 x RHA801 (Fig. 1 and 4Go–6) and an F2 population developed from an oilseed x wild (HA89 x ANN1238) hybrid (Burke et al., 2002). The number of loci producing genetic variability for seed traits obviously could be less than 54 because of pleiotropy.

The genomic regions flanking B, P, and Hyp apparently harbor multiple linked or pleiotropic QTL and seem to be hotbeds of allelic diversity for seed and other traits in sunflower. Of the seven QTL clusters identified in RHA280 x RHA801, two on the lower end of LG 17 and one on LG 10 overlapped the 100-seed weight, length, and width QTL identified by Burke et al. (2002) in HA89 x ANN1238 (Fig. 1 and Table 4). Conversely, none of the swt, sl, and sw QTL on LG 2, 3, 5, 6, 8, 9, 12, and 13 identified by Burke et al. (2002) overlapped QTL identified in RHA280 x RHA801 (Fig. 1). The P locus sits in the middle of the LG 17 QTL cluster, whereas the B locus sits in the middle of the LG 10 QTL cluster (Fig. 1). No branching QTL were identified by Burke et al. (2002) on LG 10. The 1.0-LOD support interval for a 100-seed weight QTL on LG 17 in HA89 x ANN1238 overlapped 1.0-LOD support intervals for a 100-seed weight QTL (swt17.1) and several other seed trait QTL (SOC17.1, kpr17.1, sl17.2, sw17.2, and sd17.1) on LG 17 in RHA280 x RHA801 (Fig. 1). Hence, the 23.6-cM ORS561-ORS811 interval on LG 17 harbors several QTL affecting multiple correlated and uncorrelated domestication and post- domestication traits (e.g., seed oil concentration, seed shattering, and self-pollination). Between the RHA280 x RHA801 and HA89 x ANN1238 analyses, 23 QTL have been identified for 17 traits in the ORS561-ORS811 interval on LG 17.

The segregation of branching and hull pigment loci in RHA280 x RHA801 facilitated genetic mapping of B, P, and Hyp and the assignment of B, P, and Hyp locations using the public linkage group nomenclature (Tang et al., 2002; Yu et al., 2003). B and Hyp were previously mapped by means of proprietary RFLP markers and independent linkage group nomenclatures, each differing from the public linkage group nomenclature (Leon et al., 1996; Mestries et al., 1998; Gentzbittel et al., 1999). Hull color differences distinguishing "striped" confectionery from "black" oilseed cultivars and RHA280 from RHA801 were caused by P and Hyp mutations (Table 1; Fig. 1 and 2). The allelism of P and Hyp to previously identified hull pigment mutations (Miller and Fick, 1997) is not known. P could be the phytomelanin pigment locus described by Mosjidis (1982). Hyp is the hypodermal pigment locus described by Leon et al. (1996).

The analysis of hull pigment loci in RHA280 x RHA801 was stimulated by the discovery of large-effect QTL for seed oil concentration linked to Hyp in ZENB8 x HA89 (Leon et al., 1996, 2003), and led to the discovery of multiple large-effect QTL, linked not only to Hyp, but to P (Table 4 and Fig. 1 and 4Go–6). P- and Hyp-linked QTL epistatically interacted with each other, with B-linked QTL, and with other QTL in RHA280 x RHA801 (Fig. 4 Go–6). By contrast, Hyp- linked QTL did not epistatically interact with other QTL in ZENB8 x HA89 (Leon et al., 1995, 2003). Moreover, epistatically interacting QTL for seed oil concentration have not been identified in other genetic analyses in sunflower (Leon et al., 1995, 2003; Mestries et al., 1998; Mokrani et al., 2002; Bert et al., 2003).

B, P, and Hyp either directly (pleiotropically) produce or are tightly linked to QTL producing genetic variability for seed traits in sunflower. B has long been known to profoundly affect several capitula and seed traits (Ross, 1939; Fick et al., 1974b; Dedio, 1980), and B-linked QTL have been shown to affect seed oil concentration and 100-seed weight in two high-oil x high-oil mapping populations, GH x PAC (Mestries et al., 1998) and XRQ x PSC8 (Bert et al., 2003). The B-linked seed oil QTL identified in GH x PAC and XRQ x PSC8 were similar in magnitude but opposite in sign—branching increased seed oil concentration in GH x PAC and decreased seed oil concentration in XRQ x PSC8. Mestries et al. (1998) concluded that B pleiotropically affected seed oil concentration, whereas Bert et al. (2003) concluded that seed oil concentration was affected by QTL tightly linked to B. Neither possibility can be ruled out; however, the direction of the effect of B in XRQ x PSC8 (Bert et al., 2003) runs counter to the effects of B in RHA280 x RHA801 (Fig. 3) and other genetic backgrounds (Ross, 1939; Fick et al., 1974b; Dedio, 1980; Mestries et al., 1998).

Gametic-phase linkage disequilibrium between B and linked QTL cannot be ruled out, particularly since seed weight and length QTL were identified on LG 10 in HA89 x ANN1238, a population where B was not segregating (Burke et al., 2002). We aligned LG 10 from HA89 x ANN1238 and RHA280 x RHA801 using common SSR marker loci (Tang et al., 2002; Yu et al., 2003) to ascertain whether QTL for swt and sl in the former were tightly linked to the B locus in the latter (QTL for seed oil concentration were not mapped in HA89 x ANN1238). The ORS815- ORS684 interval flanks B in RHA280 x RHA801 (Yu et al., 2003); hence, B seems to be in the middle of the 23-cM ORS878-ORS613 interval in HA89 x ANN1238, proximal to the swt and sl QTL identified by Burke et al. (2002). The latter were deduced to overlap B-linked QTL for seed traits identified in RHA280 x RHA801 (Fig. 1). The lack of branching QTL and presence of seed trait QTL in the ORS1088-ORS613 interval in HA89 x ANN1238 and large effects of B-linked QTL on seed oil concentration in opposite directions in different genetic backgrounds suggests QTL identified in the DNA marker interval flanking the B locus could have been caused by the pleiotropic effects of B and linkage disequilibrium between B and QTL (Mestries et al., 1998; Burke et al., 2002; Bert et al., 2003; Fig. 1).

The utility of exotic germplasm for enhancing the performance of elite cultivars for quantitative traits normally cannot be assessed through phenotypic analyses alone, although special phenotypic methods have been developed for identifying exotic populations and inbred lines carrying novel favorable alleles for enhancing the parents of elite single-cross hybrids (Dudley, 1984, 1987, 1988; Bernardo, 1990; Metz, 1994). QTL analyses, by contrast, are designed to identify favorable alleles in exotic and elite germplasm sources (Tanksley and Nelson, 1996; Tanksley et al., 1996; Lynch and Walsh, 1998, p. 477–489; Barton and Keightley, 2002; Doerge, 2002). One of our goals was to assess whether or not favorable alleles for seed oil concentration were present in the low-oil inbred line RHA280, an exotic, albeit narrow, source of genetic diversity for breeding high-oil sunflower. Favorable alleles were present in nearly equal numbers in the parents of three high-oil x high-oil crosses (Mestries et al., 1998; Mokrani et al., 2002; Bert et al., 2003). By contrast, Leon et al. (2003) found no favorable alleles for increasing seed oil concentration in the low-oil parent (ZENB8) of a low-oil x high-oil cross (ZENB8 x HA89), and we found no favorable alleles for increasing seed oil concentration in the low-oil parent (RHA280) of our low-oil x high-oil cross (Fig. 1 and 4Go–6 and Table 4). RHA280 and other confectionery inbred lines in the public germplasm collection (http://www.ars-grin.gov) originated from a narrow genetic base (Cheres and Knapp, 1998) and carry many of the alleles found in elite oilseed inbred lines (Tang and Knapp, 2003); hence, favorable alleles for increasing seed oil concentration and enhancing the performance of elite oilseed cultivars could be scarce in confectionery but should be common in wild sunflower germplasm (Burke et al., 2002; Gandhi et al., 2005).


   ACKNOWLEDGMENTS
 
This research was funded by grants to S.J. Knapp from the United States Department of Agriculture (USDA) National Research Initiative Competitive Grants Program Plant Genome Program (#98-35300-6166) and the USDA Cooperative State Research Education and Extension Service Initiative for Future Agricultural and Food Systems Plant Genome Program (#2000-04292).

Received for publication January 4, 2005.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 




This article has been cited by other articles:


Home page