 |  |
| ||||
Copyright © 2006 by the Genetics Society of America Dormancy Genes From Weedy Rice Respond Divergently to Seed Development Environments 1Corresponding author: Biosciences Research Laboratory, USDA-Agricultural Research Service, 1605 Albrecht Blvd., Fargo, ND 58105. E-mail: gux/at/fargo.ars.usda.gov Communicating editor: O. Savolainen Received August 3, 2005; Accepted October 18, 2005.  This article has been cited by other articles in PMC. | |||||||||||||||
Abstract Genes interacting with seed developmental environments control primary dormancy. To understand how a multigenic system evolved to adapt to the changing environments in weedy rice, we evaluated genetic components of three dormancy QTL in a synchronized nondormant genetic background. Two genetically identical populations segregating for qSD1, qSD7-1, and qSD12 were grown under greenhouse and natural conditions differing in temperature, relative humidity, and light intensity during seed development. Low temperatures tended to enhance dormancy in both conditions. However, genotypes responded to the environments divergently so that two populations displayed similar distributions for germination. Additive and/or dominance effects of the three loci explained ~90% of genetic variances and their epistases accounted for the remainder in each environment. The qSD1 and qSD7-1 main effects were increased, while the qSD12 additive effect was decreased by relatively low temperatures. Both gene main and epistatic effects were involved in G × E interactions, which in magnitude were greater than environmental main effect. The divergent responses of dormancy genes observed in this simple multigenic system presumably have selective advantages in natural populations adapted to changing environments and hence represent a genetic mechanism stabilizing the dormancy level of weedy rice ripened in different seasons or temperature regimes. | |||||||||||||||
GENES interact with each other and with environments to regulate phenotypic variation for many adaptive traits in natural populations. Environmental factors during seed development strongly influence the level of primary dormancy, and plant species or genotypes differ in their response to the environment (Bewley and Black 1982). Relatively low temperatures often enhance dormancy, which is characterized by delayed or reduced rate of germination in cereal crops (Reiner and Loch 1975; Goldbach and Michael 1976; Hayas and Hidaka 1979; Reddy et al. 1985). Genotypic difference in response to the environment affects the genetic composition of a weedy or wild population in subsequent seasons (Sawhney and Naylor 1979). Historical data for the response to local environments can be used to forecast dormancy levels of cultivars to assist in malting and breeding (Reiner and Loch 1975). Despite the broad impact on weed and crop sciences, little is known about how individual dormancy genes respond to seed development environments to regulate adaptive variation at a population level. An early genetic model for seed dormancy proposed three Mendelian factors (Johnson 1935), which demonstrated the importance of a multigenic system in controlling this adaptive trait. Subsequent research using classical genetic approaches added nonallelic interaction to the model (Jana et al. 1979, 1988; Fennimore et al. 1999; Gu et al. 2003) and emphasized environment and genotype-by-environment (G × E) interaction effects (Chang and Yen 1969; Upadhyay and Paulsen 1988; Paterson and Sorrells 1990). Recently, the Mendelian factors associated with seed dormancy were resolved as quantitative trait loci (QTL) in major cereal crops to seek dormancy genes that impart resistance to preharvest sprouting (PHS) in breeding (Anderson et al. 1993; Ullrich et al. 1993; Lin et al. 1998; Lijavetzky et al. 2000) and in wild and weedy species to determine evolutionary and genetic mechanisms underlying the adaptive trait and germination (Cai and Morishima 2000; Alonso-Blanco et al. 2003; Gu et al. 2004; Zhang et al. 2005). Dormancy QTL epistases were often detected in the QTL analyses (Anderson et al. 1993; Oberthur et al. 1995; Alonso-Blanco et al. 2003; Gu et al. 2004; Kulwal et al. 2004). However, the interlocus interactions have rarely been examined in relation to their gene additive and dominance components. Without information about the magnitude of component effects, it is difficult to understand how a multigenic system regulates the adaptive variation under changing environments and to predict the expression of selected dormancy genes in successive generations of breeding populations. Epistasis imparts an important genetic basis for the evolution of adaptation in plants (Allard 1996), but it often complicates a QTL analysis for adaptive traits and interpretation of the mapping results (Wade 2001). Different analytic systems, such as F-infinite and F2-metric models, can be used to estimate QTL epistases, and statistically each system has its own advantages (Kearsey and Pooni 1996). The F2-metric, or Cockerham's (1954) model, is considered to be more appropriate than others for modeling epistasis in a primary segregation (e.g., F2) population when the genes are in linkage equilibrium (Kao and Zeng 2002). In such a diploid population, the genetic variance can be partitioned into 3n − 1 (n is the number of loci) independent component variances using Cockerham's orthogonal contrast scales, which correspond to n additive, n dominance, and 3n − 2n − 1 epistatic effects. Cockerham's model has been used to determine the epistatic components between two QTL for maize domestication-related traits (Doebley et al. 1995). Extension of Cockerham's model from two to three loci is statistically straightforward (Cockerham 1954). However, an increase in loci also adds to the difficulty of interpretation of estimates based on data from a limited experimental population, especially when environmental factors affect gene expression (Wade 2001). Fortunately, the difficulty should be mitigated by introduction of target genes into the same genotypic background (Doebley et al. 1995). Previous research detected interactions of some dormancy QTL (e.g., qSD12) with environments (i.e., time of afterripening and year) in the weedy rice-derived primary segregation [backcross (BC)1] population (Gu et al. 2004, 2005a). Some other dormancy QTL such as qSD1 and qSD7-1 varied significantly in effect with generations of backcrossing or populations (Gu et al. 2005b). In this research, we introduced the qSD1, qSD7-1, and qSD12 dormancy alleles into a nondormant genetic background and evaluated their component genetic effects in two distinct environments during seed development. On the basis of the experimental results, we discussed: (1) to what extent epistases may contribute to the adaptive variation in a multigenic system, (2) some characteristics of dormancy G × E interactions, and (3) how dormancy genes evolved to adapt to changing environments, with emphasis on temperatures during seed development in weedy rice. | |||||||||||||||
MATERIALS AND METHODS Development of the segregation population: The donor of the dormancy genes at qSD1, qSD7-1, and qSD12 is SS18-2, an accession of wildlike weedy rice that originated from Thailand (Suh et al. 1997). The recipient parent used in the repeated backcross to transfer these dormancy genes is EM93-1, a nondormant, extremely early maturation breeding line. From the SS18-2-derived BC4F2 (132) population (Gu et al. 2005b), a plant (no. 60), which was heterozygous for qSD1, qSD7-1, and qSD12 regions, was self-pollinated to develop the trigenic segregation population. The remaining chromosomal (chr) regions, including other dormancy QTL, in the selected plant are identical to EM93-1 (Figure 1), which was determined by the markers distributed on the framework genetic map (Gu et al. 2004). It is known that the above three dormancy loci do not link with the QTL for seed shattering (Gu et al. 2005a,b), which facilitates the management of segregation populations during harvesting and reduces the possible influence of shattering on dormancy evaluation. In addition, correlation between markers on these three SS18-2-derived segments and flowering time was not detected in our experiment to isolate these three QTL as single Mendelian factors using a BC4F2 genotype similar to plant no. 60 (Gu et al. 2006). Therefore, the highly synchronized genetic background and the absence of discernable segregation for other life-history traits, which may influence germination due to tight linkage (Cai and Morishima 2000; Takeuchi et al. 2003), allow us to focus on dormancy genes underlying the three QTL.
Plant cultivation and environmental conditions: Seeds from plant no. 60 were completely afterripened and germinated, and seedlings cultured using methods previously described (Gu et al. 2004). Five weeks after germination, approximately one-half of the tillers from each plant were split; split tillers were transplanted into a new pot, and the remaining tillers were left in the original pot. The pots (28-cm diameter × 25-cm height) were filled with a mixture of clay soil and SUNSHINE medium (Sun Gro Horticulture Canada, Seba Beach, AB). A total of 234 plants were duplicated using the split-tiller technique. The two genetically identical populations were maintained in a greenhouse before flowering. Day/night temperatures were set at 29°/21°. Day length was the same as that under the local condition (Fargo, North Dakota at 46.89° north latitude and 96.79° west longitude) with supplementary light applied before 10 am and after 2 pm.Flowering date was recorded daily by tagging the first panicle emerging from the leaf sheath. The original set of plants from which tillers were taken flowered from July 5 to 18, and this population was maintained in the greenhouse during seed development. The split-tiller-derived population flowered from July 15 to 30 and was moved outdoors (~5 m away from the greenhouse) on July 17. Climatic data were automatically recorded using a HOBO Micro Station equipped with a photosynthetically active radiation (PAR) Smart Sensor (Onset, Pocasset, MA). The sensors were mounted at the plant canopy level. Mean temperature, relative humidity (RH), and PAR during the period from the onset of flowering to the end of harvest were 26.3° ± 1.8°, 70.0 ± 5.6%, and 198.8 ± 43.0 μm m−2 sec−1, respectively in the greenhouse and 22.0° ± 2.8°, 73.3 ± 8.2%, and 440.3 ± 126.9 μm m−2 sec−1, respectively, under natural conditions. Daily mean temperatures during the period are shown in Figure 2, as the data analysis below revealed their correlations with germination in each environment.
Seeds were harvested at 40 days after flowering, cleaned by removal of empty and immature spikelets, and air dried in the greenhouse for 3 days to ~12% moisture content. Dried seeds were stored at −20° to maintain dormancy. Phenotypic and genotypic identifications: Weak or moderate dormancy usually reduces germination rate, while strong dormancy delays germination. Thus, the degree of dormancy for a plant was evaluated with percentage of germination of seed samples afterripened at room temperature (24°–25°) for 10, 30, and 50 days to better display genotypic difference in a segregation population (Gu et al. 2003). About 50 seeds were placed in 9-cm petri dishes lined with a Whatman no. 1 filter paper and wetted with 10 ml deionized water. Three replications for each afterripening treatment were incubated at 30° and 100% RH in the dark for 7 days. Germination was evaluated visually by protrusion of the radicle or coleoptile from the hull by ≥3 mm.The population of 234 plants was genotyped with rice microsatellite (RM) markers on the three SS18-2-derived segments (Figure 1). Genomic DNA was prepared from young leaves. DNA was extracted, the markers were amplified by polymerase chain reaction (PCR), and the PCR products were displayed using the same methods as previously described (Gu et al. 2004). The genotyping data show that each of the 27 trigenic genotypes was replicated no less than three times in each environment (see Figure 5). Intermarker distances were adjusted with MAPMAKER/EXP 3.0 (Lincoln et al. 1992).
Data analysis and genetic parameter estimation: Germination percentage (y) for each sample was transformed by sin−1(y)−0.5, and the transformed y was averaged over three replications for further analysis. Linear correlation analysis was used to estimate the influence of climatic factors on dormancy in each environment. For this analysis, germination data from plants flowering on the same day regardless their genotypes were averaged, and the 40-day period from flowering to harvest for each plant was divided into four 10-day intervals to calculate mean temperature, RH, and PAR. Then, the mean germination was correlated with the mean temperature, RH, and PAR, respectively, to determine major climatic factors affecting dormancy and to estimate the most sensitive period of seed development in response to the environmental cues under greenhouse and natural conditions. In addition, a pairwise comparison between two genetically identical plants was used to determine mean environmental effects.One-way ANOVA was used to confirm the three QTL and their peak positions. This analysis was based on the linear model in which a phenotypic value was partitioned into the mean, genotypic, and residual (including random error and those unexplained by the other genetic effects) components. The contribution (R2) of each QTL was calculated as the proportion of the component type III sum of squares (SS) to the corrected total SS. The ANOVA was performed using the SAS procedure GLM (SAS Institute 1999). Cockerham's (1954) model was extended to three loci to partition the genetic effect and variance of these QTL into the additive, dominance, and epistatic components. Germination data from the two environments were analyzed separately on the basis of a reduced multiple linear model,
![[var epsilon]](corehtml/pmc/pmcents/epsiv.gif) ijkl is the residual including random error and trigenic epistatic effects, if any. Trigenic epistases are ignored in the model because of a relatively small sample size for some trigenic genotypes in the population.One of the advantages of Cockerham's model is that genetic variance can be partitioned into independent components, and there is no genetic covariance between components because of the property of orthogonal scales (Kao and Zeng 2002). The twenty-six orthogonal contrast scales for an F2 population segregating for loci A (a), B (b), and C (c) were developed (Table 1) to estimate individual genetic variances
Germination data from the greenhouse and natural conditions were also combined together to estimate each component G × E effect. This analysis was based on the joint model,
| |||||||||||||||
RESULTS Environmental influence and the population response: Mean germination averaged over the plants flowering on the same day significantly correlated with mean temperature in both environments and with mean RH and PAR in the natural environment (Table 2). The largest environmental influence occurred during the period from 11 to 30 days after flowering, when relatively low temperature or low PAR and high RH reduced germination, as indicated by the signs of the correlation coefficients. Temperature also correlated with RH and PAR (Table 2), suggesting that these three climatic factors did not independently influence dormancy, especially in the natural environment. Relatively, plants appeared most sensitive to temperature because the greenhouse plants responded only to variation in mean temperature from 25.8° to 26.9° (Table 2).
Mean temperature in the greenhouse (26.3°) was ~4° higher than that (22.0°) under natural conditions during seed development. Pairwise comparison between split-tiller-derived identical plants detected relatively small (<5%) differences in mean germination at 10−50 days of after ripening (DAR) (Figure 3). The largest (4.9%) and smallest (2.5%) differences occurred at 10 and 50 DAR, respectively, when the plants grown in the greenhouse had higher mean germination than those under the natural conditions. However, at 30 DAR the plants under the natural condition displayed a little (3%) higher mean germination than the greenhouse plants. This set of estimates suggests that afterripening tends to diminish the effect of seed development environment on germination and may interact with the environmental effect as dormancy is gradually released.
The two genetically identical populations displayed similar segregation patterns for germination at 10, 30, and 50 DAR (Figure 3), although they experienced distinctly different temperatures during seed development (Figure 2). Germination of seeds from the split-tiller-derived identical plants grown under greenhouse and natural conditions was correlated, with r = 0.61–0.82 (Figure 3). The correlation suggests that a common reason (mainly the identical genotypes) accounted for only part (~37–68%) of the above phenotypic similarities between two environments. Component genetic variances and effects of the three dormancy QTL: One-way ANOVA detected qSD1, qSD7-1, and qSD12 from each of the two populations (Table 3), and they are nearest to the markers RM220, RM5672, and RM270, respectively (Figure 1). These codominant markers were used to represent the respective dormancy loci in the following analyses. The three QTL contributed (R2) a relatively small, moderate, and large amount, respectively to total variances in the synchronized genetic background (Table 3). Individual QTL contributions differed between greenhouse and natural conditions and the differences varied with QTL. For example, at 10 DAR both qSD1 and qSD7-1 contributed about two times more, while qSD12 contributed approximately two times less to the phenotypic variances under natural than under greenhouse conditions. The distinct differences between the populations segregating for the same set of three dormancy QTL demonstrate that the underlying genes responded divergently to seed developmental environments.
The population of 234 plants consisted of all 27 genotypes for qSD1 (A/a), qSD7-1 (B/b), and qSD12 (C/c), with the upper- and lowercase letters standing for dormancy and nondormancy alleles at each locus, respectively (refer to Figure 5). The observed allelic frequencies are pA = 0.5171 and pa = 0.4829, pB = 0.4957 and pb = 0.5043, and pC = 0.4701 and pc = 0.5299. Ten of the 27 genotypes had sample sizes of less than five (refer to Figure 5), which is too small for the chi-square approximation for a trigenic segregation ratio. However, the observed digenic genotypic frequencies fit the expectations for two unlinked loci, i.e., (0.25:0.5:0.25) (0.25:0.5:0.25), with χ2 = 8.2 (P = 0.41) for qSD1 and qSD7-1, χ2 = 10.7 (P = 0.22) for qSD1 and qSD12, and χ2 = 4.8 (P = 0.78) for qSD7-1 and qSD12. The analysis based on Equation 1 in materials and methods detected additive effects for qSD1 (a1), qSD7-1 (a2), and qSD12 (a3), dominance effects for qSD7-1 (d2) and qSD12 (d3), and some epistases involving all aforementioned main effects and the qSD1 dominance (d1) effect (Table 4). The a1–a3 and d2 components reduced germination at 10–50 DAR under both conditions, whereas d3 increased germination and was significant only at 50 DAR. Absolute values of a1 and a2 or d2 were higher, while a3-values were lower under natural than under greenhouse conditions. The a1–a3 components totally accounted for a vast majority (80–90%) of the genetic variance, with the a3 (a3 = −0.178 to −0.376 equivalent to 3.5–13.5% germination) contributing most (42–78%) at the three DAR and in the two environments (Table 4). In contrast, d2 and d3 totally explained a relatively small amount (2–12%) of the genetic variances. It is clear that together these gene additive and dominant effects accounted for a major part of the above divergent responses of the three-locus system to the environments.
Seven sets of digenic epistasis, including additive × additive, additive × dominance, dominance × additive, and dominance × dominance, totally accounted for 1–7% of genetic variances (Table 4). Different from the above additive effects on reducing germination, the epistatic effects increased or decreased germination, which varied depending on environment and DAR. For example, four of the five epistases at 10 DAR were present under the greenhouse condition and increased germination, whereas the other one was present under the natural condition and reduced germination; the d1 × d2 and d1 × d3 epistases occurred only at 10 DAR, and the a2 × a3 and d2 × a3 occurred only at 30 and 50 DAR. Dissection of the epistases, such as the four types of qSD12-involved epistases, revealed a variety of interaction patterns (Figure 4, A–D). For example, when qSD12 was homozygous for nondormancy (cc) and dormancy (CC) alleles, the qSD1 additive effect reduced (i.e., AA < aa) and increased (i.e., AA > aa) germination, respectively (Figure 4A); similarly, when qSD12 was cc and CC, the dormancy allele at qSD7-1 was completely dominant (i.e., Bb = BB) and overdominant (i.e., Bb < BB) over the nondormancy allele, respectively (Figure 4B). With respect to diversity and their presence or absence, epistases also made an important contribution to the regulation of genetic variation with growth environment and DAR.
Genic effects estimated on the basis of Equation 1 together accounted for 94–97% of the total genetic variances in germination at different DAR and environments (Table 4). According to these estimates based on the reduced model, broad-sense heritabilities were greater under greenhouse ( Genotype-by-environment interactions: Most of the 27 trigenic genotypes varied in germination between greenhouse and natural environments (Figure 5). Some genotypes (e.g., aaBbcc, AaBbcc, and AABBCC) had lower, while others (e.g., AabbCc, AabbCC, and AaBBCC) had higher germination in the natural than in the greenhouse environment. The divergent genotypic responses on average contributed to phenotypic similarity between the populations under different environments (Figure 3) and suggest the presence of genotype-by-environmental interactions.The analysis based on Equation 3 in materials and methods corroborated the additive and dominance effects and some of the epistatic effects (Table 4) and yielded additional information about epistatic, environmental, and G × E interaction effects (Table 5). The a1 × a2 epistasis absent in model (1) was significant at 50 DAR in the combined analysis; whereas, the a1 × d3 epistasis in model (1) at 10 DAR was not significant and the d1 × d2 and d1 × d3 epistases in model (1) were shifted to G × E interactions in this analysis. Of the five gene main effects, only the qSD12 additive effect (a3) was also involved in the G × E interaction. The environmental effects (be = 0.022, −0.018, and 0.018 at 10, 30, and 50 DAR, respectively) were minor, as they were much smaller than any component gene and G × E interaction effect in absolute value (Table 5). The three sets of G × E interactions displayed divergent effects on germination and were maintained for different DAR in the population. Specifically, the a3 × E interaction ( | |||||||||||||||
DISCUSSION Contribution of epistatic effects to dormancy in a multigenic system: The synchronized genetic background improved estimation for genic effects of dormancy QTL. All three loci had significant main effects and were also involved in digenic epistases through additive and/or dominance effects (Table 4). The locus qSD1 was suggested only by interactions between its flanking marker RM259 (Figure 1) and other dormancy loci in the primary segregation (BC1) population (Gu et al. 2004). In this three-locus system, the qSD1 additive effect (a1) contributed ~0.9% (or 1.1%), while the qSD1-involved digenic epistases together contributed 4.9% (or 6.3%) to the phenotypic (or genotypic) variance in germination at 10 DAR under the greenhouse condition (Table 4). The relatively higher proportion of component epistatic variance partly explains why researchers could detect more epistatic-effect (E) than main-effect (M)-QTL in a complex genetic background (Kulwal et al. 2004) and suggests that an E-QTL for an adaptive trait is not necessarily only a regulatory locus (Wade 2001). Synchronizing the genetic background appears crucial to clarifying the nature of E-QTL and how they regulate or are regulated by an interacting gene system. For example, the simulation of a hypothesized three-locus interacting system suggests that the genetic background at two regulatory loci alters dominance performances of the third locus in its effects on fitness from neutral, additive, dominant, and over- or underdominant (Wade 2001). The dormancy genes in the present research differ from the hypothesized loci, because both qSD1 and qSD12 are basically additive and qSD7-1 is nearly completely dominant on the basis of their main effects (Table 4). However, the phenomena simulated by Wade (2001) also occurred in our three-locus system where the effect of a dormancy allele could be enhanced, offset, or inverted by a change in nonallelic combinations at the other loci (Figure 4). Gene frequency varies in weedy populations in agroecosystems because of human disturbance. Epistases make it difficult to determine which genotype(s) is a favorable nonallelic combination(s) in a population to adapt best to a particular environment. A similar difficulty also exists in breeding activities where dormancy genes are employed to improve resistance to preharvest sprouting.Digenic epistases together contributed up to 5% (or 8%) to phenotypic (or genetic) variances in the three-locus system (Table 4). Exclusion of trigenic epistases in model (1) due to limitation of sample size for some genotypes must have led to an underestimation of the epistatic contribution. This is because: (1) this model detected 94–97% or missed 3–6% of the total genetic variances, (2) higher-order epistases were detected in the BC1 population (Figures 7 and 8 in Gu et al. 2004), and (3) our simulation based on the same data and the full model, which is modified by addition of all trigenic epistases to Equation 1 (refer to Table 1), suggests the presence of several trigenic epistatic effects, such as a1 × a2 × a3 ( Partitioning genotypic means with model (1) revealed that the relatively small proportion of epistatic variation played an important role in regulating genotypic responses to seed development environments (Figure 5). For example, the AAbbCc genotypic means at 10 DAR were similar in both environments with the difference being 0.014, but their genetic components differed, including the four sets of digenic epistases (Table 6). The a1 × d3 epistasis was present in the natural, but absent under the greenhouse conditions, while the remaining three epistases were present in the greenhouse, but absent under the natural conditions. The presence or absence of epistatic effect(s) partly counteracted background (u) or gene main effects in each environment to contribute to the phenotypic similarity across the two environments (Figure 5A). Similarly, the AaBbCc and AabbCC genotypic means were 0.127 higher and 0.159 lower in the greenhouse than in the natural environments, respectively, partly because both the d1 × d2 and d1 × d3 epistatic effects increased germination in the genotype AaBbCc, but inhibited germination in the genotype AabbCC in the greenhouse environment (Table 6). These two epistatic effects together contributed 44% to the AaBbCc and 36% to the AabbCC genotypic differences between the two environments. Divergent genotypic responses to germination environments were reported for an Arabidopsis recombinant inbred line population, which led to the hypothesis that there may be different sets of genes controlling germination timing in alternative environments or that the same genes increase germination in one but decrease germination in the other environment (Donohue et al. 2005). Genetic mechanisms governing the acquiring and release of seed dormancy are likely different. Our observations demonstrate that even a simple multigenic system is capable of regulating genotypic responses through adjusting gene component effects to adapt to changing environments.
Implications of component G × E interactions for dormancy: Although interactions between genotypes and seed development environments were frequently reported in classical genetic analysis for seed dormancy and preharvest sprouting (Upadhyay and Paulsen 1988; Paterson and Sorrells 1990), usually a QTL analysis could detect the involvement only of loci with a relatively larger effect in a significant G × E interaction (Oberthur et al. 1995; Lijavetzky et al. 2000; Kulwal et al. 2004; Gu et al. 2005a). In the present research, the additive effect of the major locus qSD12 (a3) was involved in a significant G × E interaction; the other two QTL with a moderate or relatively small main effect also interacted with the environment through epistases, which were confirmed by the joint model (Table 5). It seems that no dormancy gene in a multigenic system can be completely independent of the seed development environment in the phenotypic effect on germination. However, detection of G × E's with relatively small effects appears to be dependant on the analytical method employed. For example, although the magnitude of a1 was small at 10 DAR, and the component d3 was not significant at 30 DAR according to model (1) (Table 4), we detected significant G × E interactions for a1 × E at 10 DAR (A population segregating solely for three loci may represent a statistically manageable multigenic system to examine component G × E interactions for a complex trait like dormancy. The information from the present research provides several additional insights. First, a G × E effect may increase or reduce the phenotypic effect (Table 5), which varies depending on individual genotypes (Table 6). Therefore, genetic-by-seed development environment interactions contribute directly to germination flexibility (Dekker et al. 1996) or phenotypic plasticity regardless of its active or passive responses in adaptation (van Kleunen and Fischer 2005). Second, the dominance × dominance epistases are more frequently involved in the interactions as compared with other component epistatic effects; this implies that some G × E interactions cannot be determined in a homogenous segregation population, such as the often-used recombinant inbred lines in a QTL analysis. Finally, the G × E interactions for seed dormancy can be detected during the early to midstages of afterripening, when the magnitude of a G × E interaction can be greater than the environmental main effect (be) (Table 5). Information from this research provides the basis to examine more complex genetic systems in environments with broader variation. Adaptive significance of dormancy gene divergent responses to seed development environments: Rice plants are more sensitive to temperature than to RH and light intensity in acquiring primary seed dormancy, with relatively low and high temperatures during seed development tending to enhance and reduce dormancy, respectively (Table 2). This tendency has been described for cultivars or pure lines that are different in seed dormancy (Reiner and Loch 1975; Goldbach and Michael 1976; Hayas and Hidaka 1979; Sawhney and Naylor 1979; Reddy et al. 1985). Apart from the general tendency, we observed that some genotypes (e.g., AaBbCC and AAbbCc) were relatively constant in degree of dormancy in different temperature regimes, and some others (e.g., AabbCC and AaBBCC) had weaker dormancy under relatively low temperature conditions (Figure 5). Divergently genotypic responses to seed development environment were also reported for germination timing in an Arabidopsis (Arabidopsis thaliana) recombinant inbred line population grown under different photoperiods (Munir et al. 2001; Donohue et al. 2005). Thus, a heterogeneous population may have an adaptive advantage over pure lines with respect to maintaining relative stability of seed dormancy in changing environments. Weeds often experience changing maturation environments due to human disturbance, such as change in cropping systems; dormancy as a major adaptive trait promotes survival of weed seeds in disturbed environments. Therefore, similar divergent responses of dormancy genotypes must also occur in weedy populations.Differential regulation of underlying genes is behind the divergent genotypic responses. The qSD1 and qSD7-1 loci and the qSD12 locus were up- and downregulated, respectively, by the low-temperature regime under natural conditions, with respect to their gene main effects in this highly synchronized nondormant genetic background (Tables 3 and 4). All loci in the simple multigenic system interacted with each other and with environments (Table 5), but there was no one environmental condition best suited to promote full expression of all the naturally occurring dormancy genes at a population level. The gene regulatory system seems geared to maintain dormancy homeostasis in natural populations under varying environmental conditions. The question is if homeostasis for dormancy genes has a selective advantage under natural conditions. The gene donor SS18-2 originated from Thailand (Suh et al. 1997), and its dormancy genes are presumably derived from wild rice (Oryza rufipogon) in the tropical region, as suggested by the QTL clusters or haplotypes for wild-like adaptive traits (Gu et al. 2005a,b). Weedy rice accompanies cultivated rice year-round in multiple cropping systems in tropical regions (Watanabe et al. 2000). Dormancy genes, such as that underlying qSD12, with phenotype enhanced by a relatively high temperature would be important for weed seeds ripened under seasonal humid, hot conditions to prevent immediate germination after maturation or shattering. Conversely, genes such as those at qSD1 and qSD7-1, with phenotype enhanced by low temperatures are important for weed seeds ripened under seasonal cool conditions. This set of dormancy genes was simultaneously introduced from the weedy rice SS18-2 by six generations of phenotypic selection alone for low germination extremes (Gu et al. 2005b). The cointroduction suggests that this set of dormancy genes is a favorable epistatic combination (Allard 1996) under high selection pressure. It is reasonable to believe that weedy and wild rice distributed in tropical and likely other areas have selected such a dichotomous genetic mechanism for dormancy to persist in a range of ecosystems by distributing germination over time. | |||||||||||||||
Acknowledgments We acknowledge T. Nelson, C. Kimberlin, and B. Hoffer for their technical assistance. Funding for this work was provided by the U.S. Department of Agriculture–National Research Initiative (0200668). | |||||||||||||||
References
| |||||||||||||||
