*Corresponding author.
Keywords: Dormancy, Epistasis, Molecular Markers, Cumulative Distributions
Barley (Hordeum vulgare L.) when used for malt must germinate rapidly upon imbibition. Endosperm starch and proteins must be hydrolyzed nearly to completion in 3 to 4 days. To assure rapid and complete germination, barley breeders have stringently selected against seed dormancy in germplasm intended for the malting industry. The inevitable consequence of this selection has been the development of barley varieties which are highly susceptible to pre-harvest sprouting after early fall rains or even heavy dew. Sprouted barley cannot be used by maltsters, and has diminished value as animal feed.
Barley varieties developed for animal feed have not undergone selection for the absence of dormancy. Many of the six-rowed barley varieties adapted to the western United States produce seed with variable to high levels of seed dormancy. One widely grown cultivar, Steptoe, produces seed with such impressive levels of dormancy that spilled grain from harvest may remain in fields for one or more years prior to germination. Grain producers using barley/wheat rotations in the Pacific Northwest often find volunteer Steptoe barley to be a significant weed problem in wheat fields (Ullrich et al., 1992).
The expression of seed dormancy in barley and other small grains has strong genetic and environmental components. Crossing studies of six Scandinavian barley varieties suggested that variation for dormancy in these genotypes was controlled by several recessive alleles, with no cytoplasmic effects (Burass and Skinnes, 1984). However, neither genes nor gene locations were identified in this study. Seed dormancy has been reported to be highly heritable, and selection for low or high dormancy has often proven successful in early segregating generations. There appear to be no direct associations between dormancy and other morphological or agronomic characters in barley (Burass and Aastveit, 1981).
The expression of dormancy within a genotype is conditional on many factors, including maternal environment and storage conditions. Environmental conditions during embryogenesis and seed ripening have a large impact on the development of dormancy. Cool air temperature is probably the most significant factor in promoting high levels of dormancy in barley (Buraas and Skinnes, 1984; Strand, 1989).
The most well-documented and widely practiced method of reducing seed dormancy in cereals is through postharvest storage at moderate temperatures (16C to 28C) and low seed moisture contents (8% to 16%), called "dry afterripening." The mechanism by which afterripening relieves dormancy is unknown, but may involve nonenzymatic oxidative reactions (Esashi et al., 1993; Leopold et al., 1988) or may result from turnover of products inhibiting germination (Dyer, 1993).
The objectives of this study were to use quantitative trait locus (QTL) analysis to identify chromosomal locations of genes involved in the maintenance and release of seed dormancy in segregating barley populations and to identify gene x phenotype and gene x gene interactions which modulated the expression of dormancy.
Yield trials consisting of 2 replications of 4 row plots measuring 4 meters x 1.33 meters seeded at a rate of 0.3 g/meter of row were grown at Bozeman, MT under irrigated and dryland conditions and at Pullman, WA under dryland conditions in 1991 and 1992. Intact heads were harvested by hand from each plot at visually estimated physiological maturity and placed in storage at -20C until removal for germination analysis. Heads were manually threshed and seed was permitted to postharvest ripen at 22C (+/- 3C) and 20% RH (+/- 10%) for 0, 7, 14 or 21 days prior to germination analysis at both locations in 1992. In 1991, seed was tested for germination after 7, 14 and 21 days of postharvest ripening at Bozeman and after 21 days at Pullman.
Germination tests were conducted under standard conditions (AOSA, 1988) using 50 seeds (Bozeman, 1991) or 2 replications of 100 seeds incubated on blotter papers saturated with 4mM CaCl2 in the dark at 20C. After 7 days, the number of germinated seeds were counted and expressed as a percentage of the total. See raw data.
The North American Barley Genome Mapping Project produced 295-point linkage maps (Kleinhofs et al. 1993) in each of the 150 doubled haploid lines used in this experiment which have since been expanded to include over 400 markers (see Graingenes for a current mapping dataset). Linkage maps consisting of RFLP markers, morphological markers, STS-PCR markers, RAPDs, isozymes and storage proteins were produced using MAPMAKER (Lander and Botstein, 1989). A skeletal map consisting of 191 well-distributed markers was selected for use in this experiment. Gene x phenotype interactions were identified using Mapmaker-QTL. LOD scores (log10 of the odds ratio among maximum likelihood models) were calculated according to Lander and Botstein (1989). QTL likelihood plots were created as described by Paterson et al. (1988). Analysis of variance was performed using GLM (SAS Institute, 1988a) to estimate location, time of postharvest ripening, location x time.
Following identification of QTL's, the molecular marker providing the greatest mean difference in percent dormancy was used to represent the chromosomal region of the QTL. Population distributions were determined for all two loci intereactions categories. Skewness and kurtosis estimates were determined (SAS Institute, 1988b) and cumulative pheotypic distributions of geonotypic classes were displayed (Figures 4a and 4b). Choo and Reinbergs (1982) suggested evaluating skewness and kurtosis of phenotypic distributions in doubled haploid populations to provide qualitative information regarding epistatic interactions. Genotypic means were plotted to display obvious non-additive interactions provides one simple graphic presentation (Figures 3a and 3b). Hayman and Mather (1955)suggested the measurement and quantification of epistatic interactions based on genotypic catagories. The 2-way epistatic intereactions were compared to determine potential 3- and 4-way epistatic intereactions. Classification on 2-, 3-, and 4-way epistatic interactions were then evaluated using SAS mixed model analysis (SAS Institute, 1992) where genotypes were considered fixed and lines within genotypes assumed random. This provided analysis of variance for gene and gene x gene interactions.
As is often the case with field research, our ability to visually estimate physiological maturity improved with practice. In 1991 the irrigated nursery at Bozeman was badly lodged, and mean germinations remained low throughout the period of postharvest ripening. This we believe was due to harvest of several plots prior to attainment of physiological maturity. This suggests that pulling peduncles prematurely promotes poor project performance.
Table 1 shows the single gene effects on germination means of lines grown under each environment and averaged over years and afterripening period. In every environment, germination was reduced in lines carrying the Steptoe allele at each indicated locus. The locus near PSR128 on chromosome 7 showed both a strong main effect on dormancy and also a dramatic interaction with the length of postharvest afterripening. None of the other loci produced a significant gene x time of postharvest ripening interaction.
We utilized several approaches to graphically display these intergenic interactions. The simplest method involves merely displaying the 2-locus treatment means (Figures 3a and 3b) and demonstrating differences in slope. A recently suggested modification of this approach was suggested by Dr. Gordon Lark (Univ. of Utah, presented at Plant Genome II) in which cumulative distributions of 2-locus genotypes are displayed. In the absense of epistatic interactions distributions should show periodic and predictable spacing. In the presence of epistasis, spacing of distributions should not follow additive gene effect predictions (Figures 4a and 4b).
The most complete description of epistatic interactions remains a numeric description. Through use of simple models, the population mean, location effects, time and location x time effects, and deviations from the mean due to additive gene effects and epistatic interactions may be estimated, as may their standard errors (Table 3).
One potential problem with this dataset and its interpretation may arise from the nature of the dormancy data itself. Independent of genotype, germination may never exceed 100%. Since many of the genotype x location means approached 100%, their distributions showed both kurtosis and skewness, characteristics determined by the underlying characteristics of trait measurement and independent of genetic interactions. Consequently the analysis of phenotypic distributions to estimate epistatic interactions (Choo and Reinbergs, 1982) provided us with little information of value. In order to determine whether the apparent epistasis we observed was merely an artifact of high average germination percentages, we selected environments and times in which germination percentage was relatively low for each of the four genotypic classes for the analysis of interaction of PSR128 and Amy2. The graphical representation of this analysis is shown in Figure 5. Even with relatively reduced mean germinations, the effect of allelic variation at Amy2 is masked when the Morex allele inhabits PSR128.
The dependence of one gene's effect on the allelic state at another locus (epistasis) can be used to suggest a possible ordered physiological pathway leading to the release from dormancy. The allelic state of the PSR128 locus appears to control expression of both ABG390 and Amy2 loci, indicating that it may represent a terminal step in dormancy regulation. Expression of the gene near Amy2 appears to be dependent on the allelic state of the genes near PSR128 and ABG390. These relationships suggest a pathway schematically presented in Figure 6. The gene near BCD402B appears to depend to a moderate degree upon the allelic state of the gene near ABG390.
This is the first report of epistatic control of genes involved in agronomically important traits. Both Edwards et al. (1987) and Paterson et al. (1991) determined that in their populations and traits, epistasis was of relatively little importance. Relative to dormancy in barley, epistasis appears to be significant. Whether this observation of significant epistatic interactions among QTL loci is the result of the use of completely homozygous lines (doubled haploids) and intensive replication, or is trait dependent demands further examination.
The overwhelming phenotypic impact of the gene located near PSR128 demonstrates that genes modifying quantitative trait expression need be neither small nor uniform in their effects. This study also demonstrated that while several approaches permit the visualization of epistatic interactions, quantitation of interactions remains best done through ANOVA and regression analysis.
The physiology and agronomic characteristics of seed dormancy have been studied for decades. However, our ability to more precisely characterize the genetics underlying this phenotype provides additional insights lacking in traditional genetic and physiological approaches. This kind of study allows the attachment of quantitative information about the dormancy phenotype to defined sites in the barley genome, thus taking a first step in bridging the gap between basic molecular biology and applied plant science. Likewise, information thus obtained may help develop the capabilities needed to produce small grain cultivars with appropriate and useful dormancy characteristics.
Buraas, T. and K. Aastveit. 1981. Investigations on seed dormancy in barley. pp. 554-561 In Barley Genetics IV: Proceedings of the 4th Intl. Barley Genetics Symp., Edinburgh Univ. Press.
Buraas, T. and H. Skinnes. 1984. Genetic investigations on seed dormancy in barley. Hereditas 101: 235-244.
Dyer, W.E. 1993. Dormancy-associated embryonic mRNAs and proteins in imbibing Avena fatua L. caryopses. Physiol. Plant. 88: 201-211.
Esashi, Y., M. Ogasawara, R. Gorecki, and A.C. Leopold. 1993. Possible mechanisms of afterripening in Xanthium seeds. Physiol. Plant. 87: 359-364.
Kleinhofs, A., A. Kilian, M.A. Saghai Maroof, R.M. Biyashev, P. Hayes, F.Q. Chen, N. Lapitan, A. Fenwick, T.K. Blake, V. Kanazin, E. Ananiev, L. Dahleen, D. Kudrna, J. Bollinger, S.J. Knapp, B. Liu, M. Sorrells, M. Heun, J.D. Franckowiak, D. Hoffman, R. Skadsen, B.J. Steffenson. 1993. A molecular, isozyme and morphological map of the barley genome. Theor. Appl. Genet. 86: 705-712.
Lander, E.S. and D. Botstein. 1989. Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121: 185-199.
Leopold, A.C., R. Glenister, and M.A. Cohn. 1988. Relationship between water content and afterripening in red rice. Physiol. Plant. 74: 659-662.
Muir, C.E. and R.A. Nilan. 1973. Registration of Steptoe barley. Crop Sci. 13: 770.
Rasmusson, D.C. and R.W. Wilcoxson. 1979. Registration of Morex barley. Crop Sci. 19: 293.
SAS Institute, Inc. 1988a. SAS/STAT Users Guide, releae 6.03 ed. SAS Institute, Cary, NC.
SAS Institute, Inc. 1988b. SAS Procedures Guide, release 6.03 ed. SAS Institute, Cary, NC.
SAS Institute, Inc. 1992. SAS technical report P-229, SAS/STAT Software: changes and enhancements, release 6.07 ed. SAS Institute, Cary, NC.
Strand, E. 1989. Studies on seed dormancy in small grain species. I. Barley. Nor. J. Agric. Sci. 3: 85-99.
Ullrich, S., P.M. Hayes, W.E. Dyer, T.K. Blake and J.A. Clancy. 1992. Quantitative trait locus analysis of seed dormancy in Steptoe barley. In M.K. Walker-Simmons and J.R. Reid, eds, Pre-Harvest Sprouting in Cereals 1992. Proceedings of the 6th International Symposium on Pre-Harvest Sprouting in Cereals, Coeur d'Alene, ID. American Association of Cereal Chemists, St. Paul, MN.
Table 1. Percent germination means over locations.
Data sorted by alleles within single locus markers.
Loc/Year
Amy 2 BCD 402B PSR 128 ABG 390
Bozeman - Irrigated: 1991/1992
S: 63.1 S: 61.7 S: 51.2 S: 60.7
M: 72.0 M: 71.1 M: 86.4 M: 72.8
F: 12.2 F: 12.0 F: 261 F: 22.3
P: .0005 P: .0005 P: .0000 P:.0000
Bozeman - Dryland: 1991/1992
S: 76.4 S: 74.8 S: 74.0 S: 75.8
M: 87.7 M: 86.3 M: 90.0 M: 86.2
F: 27.0 F: 24.2 F: 55.1 F: 22.6
P: .0000 P: .0000 P: .0000 P:.0000
Pullman: 1991/1992
S: 68.0 S: 66.6 S: 56.4 S: 64.8
M: 77.1 M: 75.8 M: 91.3 M: 78.6
F: 11.3 F: 10.0 F: 243 F: 26.0
P: .0008 P: .0016 P: .0000 P:.0000
*All data for each year and time of post-harvest ripening
were used to generate entry means
S: mean of lines carrying Steptoe allele
M: mean of lines carrying Morex allele
F: F value calculated by SAS, GLM
P: p value for means contrast
Table 2. Estimates and standard error of estimates for 16
possible genotypes from the Steptoe/Morex Doubled
Haploid population. Gene A = Amy2, Gene B = PSR128,
Gene C = ABG390, and Gene D = BCD402B.
Standard Error of
Genotype Estimate Estimate
AABBCCDD 54.3 +|- 2.92
AABBCCdd 62.3 +|- 3.16
AABBccDD 74.8 +|- 3.02
AABBccdd 78.5 +|- 3.08
AAbbCCDD 85.7 +|- 3.12
AAbbCCdd 95.2 +|- 3.03
AAbbccDD 98.8 +|- 2.98
AAbbccdd 97.2 +|- 3.10
aaBBCCDD 69.0 +|- 3.15
aaBBCCdd 81.9 +|- 3.15
aaBBccDD 86.7 +|- 3.00
aaBBccdd 88.4 +|- 2.92
aabbCCDD 93.0 +|- 3.23
aabbCCdd 100.6 +|- 3.05
aabbccDD 96.5 +|- 3.53
aabbccdd 99.8 +|- 3.14
Table 3. Estimates and standard error of estimates of
location, time, location x time interaction, the
mean, additive, and digenic epistatic interactions
for seed dormancy.
Standard Error of
Parameter Estimate Estimate
Location -9.4 +|- 0.98
Time 5.8 +|- 0.94
Location x Time 0.8 +|- 0.33
Mean 87.7 B +|- 3.00
Additive PSR128 -10.6 +|- 0.47
Additive ABG390 -4.9 +|- 0.46
Additive Amy2 -4.3 +|- 0.47
Additive BCD402B -2.8 +|- 0.47
Interaction PSR128, ABG390 -3.5 B +|- 0.71
Interaction Amy2, PSR128 -3.5 B +|- 0.69
Interaction ABG390, BCD402B -2.7 B +|- 0.62
Interaction Amy2, ABG390 -2.4 B +|- 0.70
B Estimates followed by the letter 'B' are biased, and
are not unique estimators of the parameters. All
estimates are significantly different from 0.00.