Research
Quantitative trait loci 9 environment interactions for plant morphology vary over ontogeny in Brassica rapa Jennifer M. Dechaine1, Marcus T. Brock2, Federico L. Iniguez-Luy3 and Cynthia Weinig2 1
Department of Biological Sciences, Central Washington University, Ellensburg, WA 98926, USA; 2Department of Botany, University of Wyoming, Laramie, WY 82071, USA;
3
Agri-Aquaculture Nutritional Genomic Center, Genetic and Bioinformatics Unit, Instituto de Investigaciones Agropecuarias-Carillanca, Codigo Postal 4780000, Temuco, Chile
Summary Author for correspondence: Jennifer M. Dechaine Tel: +1 509 963 2878 Email: [email protected] Received: 13 June 2013 Accepted: 15 August 2013
New Phytologist (2014) 201: 657–669 doi: 10.1111/nph.12520
Key words: Brassica rapa, evolutionary development, modularity, ontogeny, phenotypic plasticity, QTL 9 environment interactions, quantitative trait loci (QTLs), trade-offs.
Growth in plants occurs via the addition of repeating modules, suggesting that the genetic architecture of similar subunits may vary between earlier- and later-developing modules. These complex environment 9 ontogeny interactions are not well elucidated, as studies examining quantitative trait loci (QTLs) expression over ontogeny have not included multiple environments. Here, we characterized the genetic architecture of vegetative traits and onset of reproduction over ontogeny in recombinant inbred lines of Brassica rapa in the field and glasshouse. The magnitude of genetic variation in plasticity of seedling internodes was greater than in those produced later in ontogeny. We correspondingly detected that QTLs for seedling internode length were environment-specific, whereas later in ontogeny the majority of QTLs affected internode lengths in all treatments. The relationship between internode traits and onset of reproduction varied with environment and ontogenetic stage. This relationship was observed only in the glasshouse environment and was largely attributable to one environment-specific QTL. Our results provide the first evidence of a QTL 9 environment 9 ontogeny interaction, and provide QTL resolution for differences between early- and later-stage plasticity for stem elongation. These results also suggest potential constraints on morphological evolution in early vs later modules as a result of associations with reproductive timing.
Introduction Phenotypic plasticity is the ability of a genotype to modify phenotypic expression in response to the environment (Schmalhausen, 1949; Bradshaw, 1965). Plasticity is widely considered to be adaptive in heterogenous environments if organisms are able to match phenotypes to local conditions (Lloyd, 1984; Schmitt et al., 1995; Dudley & Schmitt, 1996; Schlichting & Pigliucci, 1998; Weinig, 2000b; Kassen, 2002; Donohue, 2003). Much of the existing literature on the evolutionary significance of plasticity focuses on one-time responses to environmental conditions experienced at a specific stage during development (Via et al., 1995; Brakefield et al., 1996, 1998; de Kroon et al., 2005). For example, in some plants, such as Arabidopsis thaliana, exposure to a given photoperiod or temperature can stimulate transition from the juvenile to the reproductive stage (Koornneef et al., 1998; Simpson & Dean, 2002). Plant vegetative growth, by contrast, occurs by the repeated addition of subunits (i.e. modules), which commonly consist of a node, subtending leaf, and an internode (section between two consecutive nodes; White, 1979; Barthelemy, 1991; Room et al., 1994; Heuret et al., 2006). Integration of traits within a module may optimize function, Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
whereas independence among vegetative modules enables a plant to respond plastically to changing environmental conditions at the level of the module (Winn, 1996a; Weinig & Delph, 2001; de Kroon et al., 2005). Variation in plasticity among modules that develop sequentially over ontogeny depends on the underlying genetic structure. If similar genetic mechanisms determine trait expression in different modules, then phenotypes are likely to be correlated between modules over ontogeny, potentially constraining the evolution of within-plant plasticity (Roach, 1986; Gomulkievicz & Kirkpatrick, 1992; Pigliucci, 1997). Alternatively, if modules are genetically uncorrelated, then those developing late in ontogeny should exhibit trait expression that is independent of earlier modules (Diggle, 1994; Pigliucci & Schlichting, 1995; de Kroon et al., 2005). Quantitative trait locus (QTL) mapping studies have identified genomic regions expressed only in one or a few developmental stages for leaf traits (Perez-Perez et al., 2002) and for consecutive internodes (Cao et al., 2001; Cheng et al., 2009; Wu et al., 2010). Each of these studies also found more modulespecific QTLs than QTLs that were conserved for the same trait measured at multiple developmental stages. These results support the hypothesis that development of morphological traits in plants New Phytologist (2014) 201: 657–669 657 www.newphytologist.com
New Phytologist
658 Research
arises primarily from module-specific allelic expression (Pigliucci & Schlichting, 1995; Winn, 1996b). This genetic architecture is likely to allow for at least some flexibility in average trait expression and plasticity among modules, although additional studies are needed to clarify the phylogenetic ubiquity of QTL 9 ontogeny interactions or possible constraints imposed by overall growth form. Phenotypic variation among modules can result from changes in within-individual plasticity to the environment and/or preprogrammed developmental changes over ontogeny (Winn, 1996b). With regard to the former, interactions between ontogeny and environment have been observed for floral sex determination (Diggle, 1994), leaf traits (Pigliucci & Schlichting, 1995; Winn, 1996a,b) and stem elongation (Dudley & Schmitt, 1995; Pigliucci & Schlichting, 1995; Pigliucci, 1997; Causin & Wulff, 2003). Specifically, species that typically encounter competition at the seedling stage exhibit plastic leaf and stem elongation responses to cues of neighbor proximity early in life (Dudley & Schmitt, 1995; Smith, 2000; Weinig, 2000a,b). Later in development, plasticity of elongation can be reduced (Weinig & Delph, 2001), either as a consequence of past selection to lessen elongation at later ontogenetic stages when neighbors cannot be overtopped and continued elongation fails to increase light interception (Dudley & Schmitt, 1996; Weinig, 2000a,b) or as a result of past selection for structural stability or improved carbon allocation that may be compromised by ongoing elongation (Givnish, 1982; Casal & Smith, 1989; Casal et al., 1994; Schmitt et al., 1995). The genetic architecture of these complex environment 9 ontogeny interactions is poorly understood, as studies examining QTL 9 ontogeny interactions have not included multiple environments (Cao et al., 2001; Perez-Perez et al., 2002; Cheng et al., 2009). Although several studies have mapped QTL 9 environment interactions for traits within a module (Weinig et al., 2002; Ungerer et al., 2003; Haselhorst et al., 2011) and floral traits (Juenger et al., 2005; Brock et al., 2010; Edwards & Weinig, 2011), few, if any, studies have identified QTLs for a single trait as expressed at different developmental stages over multiple environments. Differences in trait expression among modules may also result from developmental variation, such as the timing of reproduction and possible trade-offs with growth (Dorn & Mitchell-Olds, 1991; Pigliucci, 1997; Haselhorst et al., 2011). Many species of plants ‘bolt’ (or elongate noticeably) at reproduction, and QTLs affecting stem elongation and onset of reproduction have been identified in Brassica rapa (Edwards & Weinig, 2011; Haselhorst et al., 2011). This developmental pattern suggests that reproductive timing could be related to the expression of vegetative traits in some modules but not others. Correlations between reproduction and vegetative traits are themselves plastic, in that trade-offs between bolting date and plant size are observed under some environmental conditions and not others (Haselhorst et al., 2011). Thus, reproductive timing may be related to vegetative morphology depending on the combination of ontogenetic stage (module) in which morphology is measured and environment. Using recombinant inbred lines (RILs) of B. rapa, we examine the genetic joint architecture of reproductive timing and New Phytologist (2014) 201: 657–669 www.newphytologist.com
vegetative characters, as well as internode elongation expressed in early and later (measured as final height) modules among plants grown in field and glasshouse environments. We address two general questions: what is the correlation structure and QTL architecture for vegetative traits and bolting date within each of the field and glasshouse environments; and do genotypic correlation structure and/or QTL expression vary over ontogeny or with the combination of environment and ontogenetic stage?
Materials and Methods Study system Brassica rapa L. (Brassicaceae) is an internationally important vegetable and seed oil crop and a parent of the hybrid-origin mustard crop, Brassica napus (canola; Prakash & Hinata, 1980). Naturalized populations of B. rapa typically occur in disturbed habitats, often in close proximity to crops (Dorn & MitchellOlds, 1991; Mitchell-Olds, 1996). A segregating progeny consisting of 160 RILs was developed from a cross between two self-compatible genotypes of B. rapa: a rapid-cycling line (IMB211) and the seed oil cultivar, yellow sarson (R500; Iniguez-Luy et al., 2009). IMB211 derives from the base Wisconsin Fast PlantTM (WFP) population, which is a rapid-cycling population produced from 10 generations of selection for early flowering and rapid generation time (Williams & Hill, 1986). This selection regime was followed by seven generations of selfing and single-seed descent. The artificial selection for early flowering and rapid generation time in IMB211 is similar to that experienced by naturalized populations of this species (Dorn & Mitchell-Olds, 1991; Mitchell-Olds, 1996). The yellow sarson (R500) genotype is an agricultural cultivar that has been planted in India for at least 3000 yr (Hinata & Prakash, 1984). In comparison with IMB211, R500 delays flowering and attains substantially greater biomass. In many regards, the IMB211 and R500 parental genotypes have experienced selection similar to other weed and crop B. rapa populations, and genetic variation segregating in these RILs may resemble that segregating in wild 9 crop hybrids found in nature (Adler et al., 1993). To develop the RILs, an F1 individual resulting from the R500 9 IMB211 cross was self-fertilized, and the resulting F2s were propagated by single-seed descent to the S5 generation (Iniguez-Luy et al., 2009), resulting in 6.25% residual heterozygosity on average genome-wide. Seeds for the field and glasshouse experiments were generated by pooling all seeds from two to three replicate plants within each S5 RIL. Experimental environments and phenotypic data collection Field environment This study was part of a larger one examining morphological traits in B. rapa; the field environment has been described in detail previously (Dechaine et al., 2007). In brief, during 29–31 April 2004, seeds of 147 B. rapa RILs and the two parental lines were planted into competitive and noncompetitive treatments in an agricultural field at the University of Minnesota (St Paul, MN, USA). A nested plot design was Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
New Phytologist used, in which three plots were planted in the field and each plot was divided into three subplots of each treatment. Fifty lines were randomly assigned to each subplot, and four replicates of each line were clustered at each planting site (details are given later). This design generated three sets of four replicate plants per treatment, for a total of 12 plants per RIL 9 treatment combination (fig. 1 in Dechaine et al., 2007). In the competitive treatment, replicates of each RIL were planted in a three column 9 six row configuration with 5 cm separating each of the 18 plants and 20 cm separating adjacent plantings of RILs. Of the 18 plants within an RIL, all data were collected on the four central plants. In the noncompetitive treatment, four plants of an RIL were planted at 20 cm intervals, also 20 cm away from the adjacent RIL planting. This design with clustered replicates was chosen because of seed availability and so that competing neighbors in the competitive treatment were of the same genotype, a situation that likely arises commonly in natural populations as a result of passive seed dispersal. However, because replicates within the same subplot were grouped spatially, observed phenotypes are not statistically independent; see the section on Quantitative genetic analysis. Trait measurements Time to bolting was scored as the number of days from planting to bolting (date when buds first differentiated from the apical meristem). At bolting, hypocotyl length (from soil level to the underside of the cotyledons) and leaf length (leaf blade at its longest point) were measured; petiole length of the longest leaf was recorded 2 wk after bolting. Plants began to senesce in the first week of August and were harvested once 50% of the rosette leaves had senesced. At harvest, seedling internode length (stem length from cotyledons to the first true leaf scar) and final height (stem length from cotyledons to the tip of the primary inflorescence) were measured. We treat hypocotyl and seedling internode length as our estimates of internode elongation of early modules; the final height is the estimate of internode elongation in later modules, because seedling internode length makes up a very small percentage of total height (on average < 5%) and later developing internodes are thus the primary determinants of final height. Glasshouse environment We used a glasshouse experiment to manipulate light-quality cues of neighbor proximity independent of other factors, such as below-ground competition, that covary with density in the field. Light treatments early in development can affect later responses, and the glasshouse experiment was thus conducted over two temporally separated trials: the first exposing seedlings to light treatments, and the second exposing larger plants to light treatments following 2 wk of growth under full-sun glasshouse conditions. For the first experimental trial, on 14 March 2007, we planted 24 replicates of each of the two parental lines and 137 of the B. rapa RILs into 7.6 cm2 plastic pots, filled with Metromix 200 soil, in glasshouses at the University of Minnesota. Eight replicates per line were randomly assigned to each of three shade treatments: foliar-shade, neutralshade, and full-sun (eight plants per RIL 9 treatment combination). Shade treatments were arranged in a split-plot design Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
Research 659
consisting of 18 subplots, six per treatment. The foliar-shade treatment was created by covering a plastic PVC frame with laminated green filters (#730, Liberty Green; Lee Filters, Burbank, CA, USA). The neutral-shade and full-sun treatments were created in the same way, except using laminated white filters (#214, Full Tough Spun; Lee Filters) and clear laminate, respectively. The total red to far-red light ratio (R : FR) and photosynthetic active radiation (PAR) on an overcast day (averaged over several measurements) were, respectively, 0.6 and 174 lmol m 2 s 1 in the foliar-shade treatment, 1.1 and 178 lmol m 2 s 1 in the neutral-shade treatment, and 1.1 and 280 lmol m 2 s 1 in the full-sun treatment, as measured by an SKR 110, 660/730 nm sensor (Skye Instruments Ltd, Llandrindod Wells, UK) and an LI-250A light sensor (Li-Cor, Lincoln, NE, USA). The foliar-shade treatment is representative of the light environment a B. rapa individual would be likely to experience within an agricultural field or within an intraspecific stand similar to the competitive field environment in this study (Haselhorst et al., 2011). Phenotypic differences between the foliar-shade and neutral-shade treatments indicate an effect of light quality (R : FR), whereas differences between the neutralshade and full-sun treatments indicate sensitivity to light quantity (PAR). Bolting date was recorded for each individual. After the majority of replicates had bolted, the plants were harvested, and hypocotyl length, leaf length, petiole length, and seedling internode length were determined using the same procedures as in the field. Seeds were planted for the second experimental trial on 17 April 2007 using the same RILs and in the manner described earlier. Eighteen replicates per line were planted (six plants per RIL 9 treatment combination). Plants were again arranged in a split-plot design, in which six subplots were each divided into the three shade treatments, but in this case, shade treatments were not applied until 2 wk after germination. In early June, the plants began to senesce, and at that time we measured final height using the same methods as in the field. Quantitative genetic analysis For all traits, a restricted maximum likelihood (REML) approach (PROC MIXED; SAS, 2001) was used to partition variation attributable to RIL (VL), subplot (VB), and residual error (VR) within each of the field and glasshouse environments. The subplot term partially accounts for the clustered replicates of each RIL within a subplot in the field. Broad-sense heritibilities were calculated as VL/VP; VL is the among-RIL variance component and VP is the total phenotypic variance for a trait (sum of VL, VB, and VR). The same model, but with treatment included as a fixed effect, was used to generate lsmeans, 95% confidence intervals, and best linear unbiased predictors (BLUPs) across environments within the field (competitive and noncompetitive) and within the glasshouse (foliar-shade, neutral-shade, and full-sun). We compared between the field and glasshouse by statistically removing subplot effects for the noncompetitive (field) and full-sun (glasshouse) treatments (i.e. calculating residuals from a model including only New Phytologist (2014) 201: 657–669 www.newphytologist.com
New Phytologist
660 Research
subplot), and then using ANOVA to test the fixed effect of treatment (noncompetitive vs full-sun) and the random effects of RIL and RIL 9 treatment on each trait. For all analyses, traits were transformed using a Box–Cox procedure (Box & Cox, 1964), which greatly improved normality and homoscedasticity. BLUPs were back-transformed for the QTL analysis, because they fit the assumption of normality for QTL mapping better than untransformed traits. The parental genotypes, as well as plants that died before bolting and plants with two or fewer neighbors in the field competitive treatment, were removed from the analyses, and only RILs with data for at least three individuals per RIL 9 treatment combination were included. Final sample sizes for a trait ranged from 730 to 1160 in the field and 910 to 1400 in the glasshouse. Best linear unbiased predictors were used to estimate genetic correlations (rG) among traits (PROC CORR; SAS, 2001). Correlations were of similar sign and magnitude when compared across competitive (or light) treatments within the field (or glasshouse; Supporting Information Table S1), and only the amongtrait correlations within the noncompetitive (field) and full-sun (glasshouse) treatments are discussed further. The use of best linear unbiased predictors (BLUPs) can introduce bias to G-matrix estimation (Hadfield et al., 2010). To explore potential biases, we estimated variance-covariance matrices (and correlation matrices) using a multivariate mixed-model approach (PROC MIXED; Brock et al., 2010; Dmitriew et al., 2010; Hadfield et al., 2010) and compared these estimates to those calculated using genotypic BLUP values (see Supporting Information for a description of additional analyses of genotypic values). We used Flury’s hierarchical common principal components (CPC) analysis, using the ‘jump-up approach’, to test for differences in matrix structure across environments (Flury, 1998; Phillips & Arnold, 1999; Mezey & Houle, 2003). This approach distills G-matrices into eigenvectors and eigenvalues and then tests hierarchically structured hypotheses across environments. The preceding G-matrix comparisons explore how multivariate relationships among vegetative and phenological traits differ across environments. To identify specific matrix elements that may contribute to G-matrix differences, we utilized Fisher’s Z-test to compare bivariate correlations across environments (see Methods S1 for an additional description of G-matrix analyses). Across-environmental correlations (rGE) among all treatment combinations within the field or glasshouse and among the field, noncompetitive and glasshouse, full-sun treatments were calculated as cov(1,2)/√(V1,1 9 V1,2), where cov(1,2) is the covariance of a trait across two environments, and V1,1 and V1,2 are the amongRIL variances within each environment (Robertson, 1959; Gurganus et al., 1998). We tested the significance of rGE estimates between the noncompetitive and full-sun treatments using the mixed-model ANOVA described earlier, testing the fixed effect of treatment (noncompetitive vs full-sun) and the random effects of RIL and RIL 9 treatment on each trait. A significant RIL term and a nonsignificant RIL 9 treatment interaction indicate that rGE does not differ significantly from 1, suggesting that loci regulating a specific trait have a similar effect across environments. A significant RIL 9 treatment interaction New Phytologist (2014) 201: 657–669 www.newphytologist.com
indicates at least some difference in rank order (or variance) among RILs across environments (i.e. rGE < 1) and that some (RIL term also significant) or a majority (RIL term not significant) of the loci regulating a trait are environment-specific (Gurganus et al., 1998; Vieira et al., 2000; Edwards & Weinig, 2011). QTL analysis The R500 9 IMB211 RILs have been genotyped at 227 markers across 10 linkage groups, representing the 10 chromosomes in B. rapa (Iniguez-Luy et al., 2009). Marker order for the linkage map was estimated from recombination frequencies observed in the entire population of 160 RILs, that is, from the most complete dataset in earlier experiments. Only 152 RILs survived to bolting in either the field or glasshouse experiments; we thus re-estimated cM distances in these 152 RILs to account for slight differences that might exist as a result of unsampled recombination events. The map distances for each marker locus were calculated from the estimated recombination frequencies using the Kosambi mapping function in R/qtl (R Development Core Team, 2007). Quantitative trait loci were mapped using the composite interval mapping (CIM) procedure in QTL Cartographer (Wang et al., 2007). To control for effects of variation segregating elsewhere in the genome, we identified cofactors using forwardbackward regression, and a 5 cM window; a maximum of five cofactors was selected for inclusion in the mapping model. Additive effects and the genetic variance explained by a QTL were calculated in QTL Cartographer and confirmed using GLM (PROC GLM; SAS, 2001) with all QTLs detected in the genome-wide screen as main effects (Lynch & Walsh, 1998). For a given locus, a positive additive effect indicates that the IMB211 parent allele conferred a higher value for the trait. We used the QTL Cartographer QTL 9 environment hypothesis test to evaluate environmental interactions. Single-marker ANOVA (Lynch & Walsh, 1998) was used to confirm QTL 9 environment interactions between the field (noncompetitive) and glasshouse (full-sun) environments; all significant main-effect QTLs for the trait, treatment, and QTL 9 treatment effects on each genotypic trait mean were included in the model (PROC GLM). A similar model was used to test for QTL 9 ontogeny interactions between QTLs affecting internode elongation at early (seedling internode length) vs later (final height) stage modules within the glasshouse or field. We focus on the quantitative-genetic and QTL architecture of the noncompetitive treatment in the field and the full-sun treatment in the glasshouse, because heritabilities were higher in these two environments than in other glasshouse (foliarshade or neutral-shade) or field (competitive) environments and, correspondingly, more QTLs were mapped. In addition, the between-environment correlations for a trait across any two light treatments (glasshouse) or across competitive treatments (field) were strongly positive (range, r = 0.71–0.99; Table S1), suggesting little change would be observed in QTL architecture for individual traits across these treatments. Finally, comparing results between the glasshouse and the field is of general Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
New Phytologist
Research 661
interest, because it bears on our ability to generalize across controlled and natural growth settings. (For results across competitive treatments within the field or across shade treatments within the glasshouse, see Tables S1–S4.)
internode length (P = 0.99), suggesting that some loci affect phenotypic expression of traits (other than seedling internode length) across both environments (Table S5). On the other hand, all estimates of rGE differed significantly from 1 (i.e. P < 0.001 for all
Results
Table 1 Quantitative genetic partitioning of variance components for plant characters in Brassica rapa recombinant inbred lines (RILs) in the field, noncompetitive (F) and glasshouse, full-sun (G) environments
Trait
F
G
F
G
F
G
Time to bolting Hypocotyl length Petiole length Leaf length Seedling internode length Final height
0.79 0.16 0.07 0.64 0.09 6.97
0.00 2.53 0.34 0.02 0.04 200.50
5.07 0.72 0.23 1.55 0.84 9.03
0.00 2.47 0.53 0.03 0.03 200.06
0.13 0.17 0.23 0.28 0.10 0.41
0.71 0.36 0.32 0.30 0.54 0.48
(45)
0
Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
Petiole length (mm) 5.0 25.0 45.0 65.0
0
50 Glasshouse
(173)
(121)
50
50
0 Field
Glasshouse
Final height (mm) 300.0 600.0 900.0
Glasshouse
50 Glasshouse
Field
50
0
0
50
Glasshouse
Field First Internode length (mm) 2.0 6.0 12.0 18.0
50
0 Field
50
50 Field
Hypocotyl length (mm) 15.0 25.0 35.0 45.0
Hypocotyl length (mm) 3.0 6.0 9.0 12.0
50
50
(40)
Glasshouse
Leaf length (mm) 10.0 40.0 70.0 110.0
50
Days to bolting 16.0 22.0 28.0 34.0
Restricted maximum likelihood estimates of the among-RIL variance component (VL), the residual variance component (VR), and the broadsense heritability (H2), calculated as VL/(VP); VP is the total phenotypic variance component.
Field
Fig. 1 Frequency plots of Brassica rapa recombinant inbred lines (RILs) best-linear unbiased predictors (BLUPs) in the field, noncompetitive and glasshouse, full-sun environments. Means of the RILs (solid, arrow) and the parental phenotypes for IMB211 (dotted, diamond) and R500 (dashed, oval) are shown. Ranges for time to bolting and hypocotyl length did not overlap for the field and glasshouse; these traits are thus shown on different scales for each environment. R500 means were outside the histogram for time to bolting and leaf length; values are shown. Petiole length was not recorded for IMB211 in the field.
H2 (VL/VP)
VR
VL
50
50
0 Field
Glasshouse
0.0
37.5
Days to bolting 39.5 41.5 43.5
Recombinant inbred lines differed significantly (P < 0.001) for the average expression of each trait within the field, noncompetitive and glasshouse, full-sun environments. The RIL mean was intermediate to both parental means for all traits except days to bolting in the field; we also observed transgressive segregation for most traits, in that some RILs exhibited trait values outside the range of the parental genotypes (Fig. 1). Broad-sense heritabilities (H2) were greater in the glasshouse, full-sun treatment (range, 0.30–0.71) than in the field, noncompetitive setting (range, 0.10–0.41) for all traits (Table 1). Across-environment genotypic correlations were positive for all traits (Table 2). Estimates of rGE differed significantly (P < 0.01 for RIL effect) or marginally significantly (P = 0.07 for RIL effect for leaf length) from zero for all traits except seedling
50
50
0 Field
Glasshouse
New Phytologist (2014) 201: 657–669 www.newphytologist.com
New Phytologist
662 Research
Table 2 Genotypic correlations among traits in Brassica rapa recombinant inbred lines in the field, noncompetitive (above diagonal) and glasshouse, full-sun (below diagonal) environments
Time to bolting Hypocotyl length Petiole length Leaf length Seedling internode length Final height
Time to bolting
Hypocotyl length
Petiole length
Leaf length
Seedling internode length
Final height
0.56*** (0.37)**** 0.52**** 0.45**** (0.29) *** 0.57****
(0.20)* 0.47**** (0.08) 0.06 0.72**** 0.17*
0.16 0.01 0.43** 0.85**** (0.15) 0.51****
0.07 0.05 0.68**** 0.61^ 0.04 0.61****
(0.03) 0.10 0.29** 0.23** 0.15 0.24**
0.06 0.14 0.61**** 0.79**** 0.50**** 0.64****
Cross-environmental correlations for each trait are on the diagonal (italicized). Pearson correlation coefficients are shown; parentheses indicate negative correlations. Significance is as follows: ^
Quantitative trait loci 9 environment interactions for plant morphology vary over ontogeny in Brassica rapa Jennifer M. Dechaine1, Marcus T. Brock2, Federico L. Iniguez-Luy3 and Cynthia Weinig2 1
Department of Biological Sciences, Central Washington University, Ellensburg, WA 98926, USA; 2Department of Botany, University of Wyoming, Laramie, WY 82071, USA;
3
Agri-Aquaculture Nutritional Genomic Center, Genetic and Bioinformatics Unit, Instituto de Investigaciones Agropecuarias-Carillanca, Codigo Postal 4780000, Temuco, Chile
Summary Author for correspondence: Jennifer M. Dechaine Tel: +1 509 963 2878 Email: [email protected] Received: 13 June 2013 Accepted: 15 August 2013
New Phytologist (2014) 201: 657–669 doi: 10.1111/nph.12520
Key words: Brassica rapa, evolutionary development, modularity, ontogeny, phenotypic plasticity, QTL 9 environment interactions, quantitative trait loci (QTLs), trade-offs.
Growth in plants occurs via the addition of repeating modules, suggesting that the genetic architecture of similar subunits may vary between earlier- and later-developing modules. These complex environment 9 ontogeny interactions are not well elucidated, as studies examining quantitative trait loci (QTLs) expression over ontogeny have not included multiple environments. Here, we characterized the genetic architecture of vegetative traits and onset of reproduction over ontogeny in recombinant inbred lines of Brassica rapa in the field and glasshouse. The magnitude of genetic variation in plasticity of seedling internodes was greater than in those produced later in ontogeny. We correspondingly detected that QTLs for seedling internode length were environment-specific, whereas later in ontogeny the majority of QTLs affected internode lengths in all treatments. The relationship between internode traits and onset of reproduction varied with environment and ontogenetic stage. This relationship was observed only in the glasshouse environment and was largely attributable to one environment-specific QTL. Our results provide the first evidence of a QTL 9 environment 9 ontogeny interaction, and provide QTL resolution for differences between early- and later-stage plasticity for stem elongation. These results also suggest potential constraints on morphological evolution in early vs later modules as a result of associations with reproductive timing.
Introduction Phenotypic plasticity is the ability of a genotype to modify phenotypic expression in response to the environment (Schmalhausen, 1949; Bradshaw, 1965). Plasticity is widely considered to be adaptive in heterogenous environments if organisms are able to match phenotypes to local conditions (Lloyd, 1984; Schmitt et al., 1995; Dudley & Schmitt, 1996; Schlichting & Pigliucci, 1998; Weinig, 2000b; Kassen, 2002; Donohue, 2003). Much of the existing literature on the evolutionary significance of plasticity focuses on one-time responses to environmental conditions experienced at a specific stage during development (Via et al., 1995; Brakefield et al., 1996, 1998; de Kroon et al., 2005). For example, in some plants, such as Arabidopsis thaliana, exposure to a given photoperiod or temperature can stimulate transition from the juvenile to the reproductive stage (Koornneef et al., 1998; Simpson & Dean, 2002). Plant vegetative growth, by contrast, occurs by the repeated addition of subunits (i.e. modules), which commonly consist of a node, subtending leaf, and an internode (section between two consecutive nodes; White, 1979; Barthelemy, 1991; Room et al., 1994; Heuret et al., 2006). Integration of traits within a module may optimize function, Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
whereas independence among vegetative modules enables a plant to respond plastically to changing environmental conditions at the level of the module (Winn, 1996a; Weinig & Delph, 2001; de Kroon et al., 2005). Variation in plasticity among modules that develop sequentially over ontogeny depends on the underlying genetic structure. If similar genetic mechanisms determine trait expression in different modules, then phenotypes are likely to be correlated between modules over ontogeny, potentially constraining the evolution of within-plant plasticity (Roach, 1986; Gomulkievicz & Kirkpatrick, 1992; Pigliucci, 1997). Alternatively, if modules are genetically uncorrelated, then those developing late in ontogeny should exhibit trait expression that is independent of earlier modules (Diggle, 1994; Pigliucci & Schlichting, 1995; de Kroon et al., 2005). Quantitative trait locus (QTL) mapping studies have identified genomic regions expressed only in one or a few developmental stages for leaf traits (Perez-Perez et al., 2002) and for consecutive internodes (Cao et al., 2001; Cheng et al., 2009; Wu et al., 2010). Each of these studies also found more modulespecific QTLs than QTLs that were conserved for the same trait measured at multiple developmental stages. These results support the hypothesis that development of morphological traits in plants New Phytologist (2014) 201: 657–669 657 www.newphytologist.com
New Phytologist
658 Research
arises primarily from module-specific allelic expression (Pigliucci & Schlichting, 1995; Winn, 1996b). This genetic architecture is likely to allow for at least some flexibility in average trait expression and plasticity among modules, although additional studies are needed to clarify the phylogenetic ubiquity of QTL 9 ontogeny interactions or possible constraints imposed by overall growth form. Phenotypic variation among modules can result from changes in within-individual plasticity to the environment and/or preprogrammed developmental changes over ontogeny (Winn, 1996b). With regard to the former, interactions between ontogeny and environment have been observed for floral sex determination (Diggle, 1994), leaf traits (Pigliucci & Schlichting, 1995; Winn, 1996a,b) and stem elongation (Dudley & Schmitt, 1995; Pigliucci & Schlichting, 1995; Pigliucci, 1997; Causin & Wulff, 2003). Specifically, species that typically encounter competition at the seedling stage exhibit plastic leaf and stem elongation responses to cues of neighbor proximity early in life (Dudley & Schmitt, 1995; Smith, 2000; Weinig, 2000a,b). Later in development, plasticity of elongation can be reduced (Weinig & Delph, 2001), either as a consequence of past selection to lessen elongation at later ontogenetic stages when neighbors cannot be overtopped and continued elongation fails to increase light interception (Dudley & Schmitt, 1996; Weinig, 2000a,b) or as a result of past selection for structural stability or improved carbon allocation that may be compromised by ongoing elongation (Givnish, 1982; Casal & Smith, 1989; Casal et al., 1994; Schmitt et al., 1995). The genetic architecture of these complex environment 9 ontogeny interactions is poorly understood, as studies examining QTL 9 ontogeny interactions have not included multiple environments (Cao et al., 2001; Perez-Perez et al., 2002; Cheng et al., 2009). Although several studies have mapped QTL 9 environment interactions for traits within a module (Weinig et al., 2002; Ungerer et al., 2003; Haselhorst et al., 2011) and floral traits (Juenger et al., 2005; Brock et al., 2010; Edwards & Weinig, 2011), few, if any, studies have identified QTLs for a single trait as expressed at different developmental stages over multiple environments. Differences in trait expression among modules may also result from developmental variation, such as the timing of reproduction and possible trade-offs with growth (Dorn & Mitchell-Olds, 1991; Pigliucci, 1997; Haselhorst et al., 2011). Many species of plants ‘bolt’ (or elongate noticeably) at reproduction, and QTLs affecting stem elongation and onset of reproduction have been identified in Brassica rapa (Edwards & Weinig, 2011; Haselhorst et al., 2011). This developmental pattern suggests that reproductive timing could be related to the expression of vegetative traits in some modules but not others. Correlations between reproduction and vegetative traits are themselves plastic, in that trade-offs between bolting date and plant size are observed under some environmental conditions and not others (Haselhorst et al., 2011). Thus, reproductive timing may be related to vegetative morphology depending on the combination of ontogenetic stage (module) in which morphology is measured and environment. Using recombinant inbred lines (RILs) of B. rapa, we examine the genetic joint architecture of reproductive timing and New Phytologist (2014) 201: 657–669 www.newphytologist.com
vegetative characters, as well as internode elongation expressed in early and later (measured as final height) modules among plants grown in field and glasshouse environments. We address two general questions: what is the correlation structure and QTL architecture for vegetative traits and bolting date within each of the field and glasshouse environments; and do genotypic correlation structure and/or QTL expression vary over ontogeny or with the combination of environment and ontogenetic stage?
Materials and Methods Study system Brassica rapa L. (Brassicaceae) is an internationally important vegetable and seed oil crop and a parent of the hybrid-origin mustard crop, Brassica napus (canola; Prakash & Hinata, 1980). Naturalized populations of B. rapa typically occur in disturbed habitats, often in close proximity to crops (Dorn & MitchellOlds, 1991; Mitchell-Olds, 1996). A segregating progeny consisting of 160 RILs was developed from a cross between two self-compatible genotypes of B. rapa: a rapid-cycling line (IMB211) and the seed oil cultivar, yellow sarson (R500; Iniguez-Luy et al., 2009). IMB211 derives from the base Wisconsin Fast PlantTM (WFP) population, which is a rapid-cycling population produced from 10 generations of selection for early flowering and rapid generation time (Williams & Hill, 1986). This selection regime was followed by seven generations of selfing and single-seed descent. The artificial selection for early flowering and rapid generation time in IMB211 is similar to that experienced by naturalized populations of this species (Dorn & Mitchell-Olds, 1991; Mitchell-Olds, 1996). The yellow sarson (R500) genotype is an agricultural cultivar that has been planted in India for at least 3000 yr (Hinata & Prakash, 1984). In comparison with IMB211, R500 delays flowering and attains substantially greater biomass. In many regards, the IMB211 and R500 parental genotypes have experienced selection similar to other weed and crop B. rapa populations, and genetic variation segregating in these RILs may resemble that segregating in wild 9 crop hybrids found in nature (Adler et al., 1993). To develop the RILs, an F1 individual resulting from the R500 9 IMB211 cross was self-fertilized, and the resulting F2s were propagated by single-seed descent to the S5 generation (Iniguez-Luy et al., 2009), resulting in 6.25% residual heterozygosity on average genome-wide. Seeds for the field and glasshouse experiments were generated by pooling all seeds from two to three replicate plants within each S5 RIL. Experimental environments and phenotypic data collection Field environment This study was part of a larger one examining morphological traits in B. rapa; the field environment has been described in detail previously (Dechaine et al., 2007). In brief, during 29–31 April 2004, seeds of 147 B. rapa RILs and the two parental lines were planted into competitive and noncompetitive treatments in an agricultural field at the University of Minnesota (St Paul, MN, USA). A nested plot design was Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
New Phytologist used, in which three plots were planted in the field and each plot was divided into three subplots of each treatment. Fifty lines were randomly assigned to each subplot, and four replicates of each line were clustered at each planting site (details are given later). This design generated three sets of four replicate plants per treatment, for a total of 12 plants per RIL 9 treatment combination (fig. 1 in Dechaine et al., 2007). In the competitive treatment, replicates of each RIL were planted in a three column 9 six row configuration with 5 cm separating each of the 18 plants and 20 cm separating adjacent plantings of RILs. Of the 18 plants within an RIL, all data were collected on the four central plants. In the noncompetitive treatment, four plants of an RIL were planted at 20 cm intervals, also 20 cm away from the adjacent RIL planting. This design with clustered replicates was chosen because of seed availability and so that competing neighbors in the competitive treatment were of the same genotype, a situation that likely arises commonly in natural populations as a result of passive seed dispersal. However, because replicates within the same subplot were grouped spatially, observed phenotypes are not statistically independent; see the section on Quantitative genetic analysis. Trait measurements Time to bolting was scored as the number of days from planting to bolting (date when buds first differentiated from the apical meristem). At bolting, hypocotyl length (from soil level to the underside of the cotyledons) and leaf length (leaf blade at its longest point) were measured; petiole length of the longest leaf was recorded 2 wk after bolting. Plants began to senesce in the first week of August and were harvested once 50% of the rosette leaves had senesced. At harvest, seedling internode length (stem length from cotyledons to the first true leaf scar) and final height (stem length from cotyledons to the tip of the primary inflorescence) were measured. We treat hypocotyl and seedling internode length as our estimates of internode elongation of early modules; the final height is the estimate of internode elongation in later modules, because seedling internode length makes up a very small percentage of total height (on average < 5%) and later developing internodes are thus the primary determinants of final height. Glasshouse environment We used a glasshouse experiment to manipulate light-quality cues of neighbor proximity independent of other factors, such as below-ground competition, that covary with density in the field. Light treatments early in development can affect later responses, and the glasshouse experiment was thus conducted over two temporally separated trials: the first exposing seedlings to light treatments, and the second exposing larger plants to light treatments following 2 wk of growth under full-sun glasshouse conditions. For the first experimental trial, on 14 March 2007, we planted 24 replicates of each of the two parental lines and 137 of the B. rapa RILs into 7.6 cm2 plastic pots, filled with Metromix 200 soil, in glasshouses at the University of Minnesota. Eight replicates per line were randomly assigned to each of three shade treatments: foliar-shade, neutralshade, and full-sun (eight plants per RIL 9 treatment combination). Shade treatments were arranged in a split-plot design Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
Research 659
consisting of 18 subplots, six per treatment. The foliar-shade treatment was created by covering a plastic PVC frame with laminated green filters (#730, Liberty Green; Lee Filters, Burbank, CA, USA). The neutral-shade and full-sun treatments were created in the same way, except using laminated white filters (#214, Full Tough Spun; Lee Filters) and clear laminate, respectively. The total red to far-red light ratio (R : FR) and photosynthetic active radiation (PAR) on an overcast day (averaged over several measurements) were, respectively, 0.6 and 174 lmol m 2 s 1 in the foliar-shade treatment, 1.1 and 178 lmol m 2 s 1 in the neutral-shade treatment, and 1.1 and 280 lmol m 2 s 1 in the full-sun treatment, as measured by an SKR 110, 660/730 nm sensor (Skye Instruments Ltd, Llandrindod Wells, UK) and an LI-250A light sensor (Li-Cor, Lincoln, NE, USA). The foliar-shade treatment is representative of the light environment a B. rapa individual would be likely to experience within an agricultural field or within an intraspecific stand similar to the competitive field environment in this study (Haselhorst et al., 2011). Phenotypic differences between the foliar-shade and neutral-shade treatments indicate an effect of light quality (R : FR), whereas differences between the neutralshade and full-sun treatments indicate sensitivity to light quantity (PAR). Bolting date was recorded for each individual. After the majority of replicates had bolted, the plants were harvested, and hypocotyl length, leaf length, petiole length, and seedling internode length were determined using the same procedures as in the field. Seeds were planted for the second experimental trial on 17 April 2007 using the same RILs and in the manner described earlier. Eighteen replicates per line were planted (six plants per RIL 9 treatment combination). Plants were again arranged in a split-plot design, in which six subplots were each divided into the three shade treatments, but in this case, shade treatments were not applied until 2 wk after germination. In early June, the plants began to senesce, and at that time we measured final height using the same methods as in the field. Quantitative genetic analysis For all traits, a restricted maximum likelihood (REML) approach (PROC MIXED; SAS, 2001) was used to partition variation attributable to RIL (VL), subplot (VB), and residual error (VR) within each of the field and glasshouse environments. The subplot term partially accounts for the clustered replicates of each RIL within a subplot in the field. Broad-sense heritibilities were calculated as VL/VP; VL is the among-RIL variance component and VP is the total phenotypic variance for a trait (sum of VL, VB, and VR). The same model, but with treatment included as a fixed effect, was used to generate lsmeans, 95% confidence intervals, and best linear unbiased predictors (BLUPs) across environments within the field (competitive and noncompetitive) and within the glasshouse (foliar-shade, neutral-shade, and full-sun). We compared between the field and glasshouse by statistically removing subplot effects for the noncompetitive (field) and full-sun (glasshouse) treatments (i.e. calculating residuals from a model including only New Phytologist (2014) 201: 657–669 www.newphytologist.com
New Phytologist
660 Research
subplot), and then using ANOVA to test the fixed effect of treatment (noncompetitive vs full-sun) and the random effects of RIL and RIL 9 treatment on each trait. For all analyses, traits were transformed using a Box–Cox procedure (Box & Cox, 1964), which greatly improved normality and homoscedasticity. BLUPs were back-transformed for the QTL analysis, because they fit the assumption of normality for QTL mapping better than untransformed traits. The parental genotypes, as well as plants that died before bolting and plants with two or fewer neighbors in the field competitive treatment, were removed from the analyses, and only RILs with data for at least three individuals per RIL 9 treatment combination were included. Final sample sizes for a trait ranged from 730 to 1160 in the field and 910 to 1400 in the glasshouse. Best linear unbiased predictors were used to estimate genetic correlations (rG) among traits (PROC CORR; SAS, 2001). Correlations were of similar sign and magnitude when compared across competitive (or light) treatments within the field (or glasshouse; Supporting Information Table S1), and only the amongtrait correlations within the noncompetitive (field) and full-sun (glasshouse) treatments are discussed further. The use of best linear unbiased predictors (BLUPs) can introduce bias to G-matrix estimation (Hadfield et al., 2010). To explore potential biases, we estimated variance-covariance matrices (and correlation matrices) using a multivariate mixed-model approach (PROC MIXED; Brock et al., 2010; Dmitriew et al., 2010; Hadfield et al., 2010) and compared these estimates to those calculated using genotypic BLUP values (see Supporting Information for a description of additional analyses of genotypic values). We used Flury’s hierarchical common principal components (CPC) analysis, using the ‘jump-up approach’, to test for differences in matrix structure across environments (Flury, 1998; Phillips & Arnold, 1999; Mezey & Houle, 2003). This approach distills G-matrices into eigenvectors and eigenvalues and then tests hierarchically structured hypotheses across environments. The preceding G-matrix comparisons explore how multivariate relationships among vegetative and phenological traits differ across environments. To identify specific matrix elements that may contribute to G-matrix differences, we utilized Fisher’s Z-test to compare bivariate correlations across environments (see Methods S1 for an additional description of G-matrix analyses). Across-environmental correlations (rGE) among all treatment combinations within the field or glasshouse and among the field, noncompetitive and glasshouse, full-sun treatments were calculated as cov(1,2)/√(V1,1 9 V1,2), where cov(1,2) is the covariance of a trait across two environments, and V1,1 and V1,2 are the amongRIL variances within each environment (Robertson, 1959; Gurganus et al., 1998). We tested the significance of rGE estimates between the noncompetitive and full-sun treatments using the mixed-model ANOVA described earlier, testing the fixed effect of treatment (noncompetitive vs full-sun) and the random effects of RIL and RIL 9 treatment on each trait. A significant RIL term and a nonsignificant RIL 9 treatment interaction indicate that rGE does not differ significantly from 1, suggesting that loci regulating a specific trait have a similar effect across environments. A significant RIL 9 treatment interaction New Phytologist (2014) 201: 657–669 www.newphytologist.com
indicates at least some difference in rank order (or variance) among RILs across environments (i.e. rGE < 1) and that some (RIL term also significant) or a majority (RIL term not significant) of the loci regulating a trait are environment-specific (Gurganus et al., 1998; Vieira et al., 2000; Edwards & Weinig, 2011). QTL analysis The R500 9 IMB211 RILs have been genotyped at 227 markers across 10 linkage groups, representing the 10 chromosomes in B. rapa (Iniguez-Luy et al., 2009). Marker order for the linkage map was estimated from recombination frequencies observed in the entire population of 160 RILs, that is, from the most complete dataset in earlier experiments. Only 152 RILs survived to bolting in either the field or glasshouse experiments; we thus re-estimated cM distances in these 152 RILs to account for slight differences that might exist as a result of unsampled recombination events. The map distances for each marker locus were calculated from the estimated recombination frequencies using the Kosambi mapping function in R/qtl (R Development Core Team, 2007). Quantitative trait loci were mapped using the composite interval mapping (CIM) procedure in QTL Cartographer (Wang et al., 2007). To control for effects of variation segregating elsewhere in the genome, we identified cofactors using forwardbackward regression, and a 5 cM window; a maximum of five cofactors was selected for inclusion in the mapping model. Additive effects and the genetic variance explained by a QTL were calculated in QTL Cartographer and confirmed using GLM (PROC GLM; SAS, 2001) with all QTLs detected in the genome-wide screen as main effects (Lynch & Walsh, 1998). For a given locus, a positive additive effect indicates that the IMB211 parent allele conferred a higher value for the trait. We used the QTL Cartographer QTL 9 environment hypothesis test to evaluate environmental interactions. Single-marker ANOVA (Lynch & Walsh, 1998) was used to confirm QTL 9 environment interactions between the field (noncompetitive) and glasshouse (full-sun) environments; all significant main-effect QTLs for the trait, treatment, and QTL 9 treatment effects on each genotypic trait mean were included in the model (PROC GLM). A similar model was used to test for QTL 9 ontogeny interactions between QTLs affecting internode elongation at early (seedling internode length) vs later (final height) stage modules within the glasshouse or field. We focus on the quantitative-genetic and QTL architecture of the noncompetitive treatment in the field and the full-sun treatment in the glasshouse, because heritabilities were higher in these two environments than in other glasshouse (foliarshade or neutral-shade) or field (competitive) environments and, correspondingly, more QTLs were mapped. In addition, the between-environment correlations for a trait across any two light treatments (glasshouse) or across competitive treatments (field) were strongly positive (range, r = 0.71–0.99; Table S1), suggesting little change would be observed in QTL architecture for individual traits across these treatments. Finally, comparing results between the glasshouse and the field is of general Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
New Phytologist
Research 661
interest, because it bears on our ability to generalize across controlled and natural growth settings. (For results across competitive treatments within the field or across shade treatments within the glasshouse, see Tables S1–S4.)
internode length (P = 0.99), suggesting that some loci affect phenotypic expression of traits (other than seedling internode length) across both environments (Table S5). On the other hand, all estimates of rGE differed significantly from 1 (i.e. P < 0.001 for all
Results
Table 1 Quantitative genetic partitioning of variance components for plant characters in Brassica rapa recombinant inbred lines (RILs) in the field, noncompetitive (F) and glasshouse, full-sun (G) environments
Trait
F
G
F
G
F
G
Time to bolting Hypocotyl length Petiole length Leaf length Seedling internode length Final height
0.79 0.16 0.07 0.64 0.09 6.97
0.00 2.53 0.34 0.02 0.04 200.50
5.07 0.72 0.23 1.55 0.84 9.03
0.00 2.47 0.53 0.03 0.03 200.06
0.13 0.17 0.23 0.28 0.10 0.41
0.71 0.36 0.32 0.30 0.54 0.48
(45)
0
Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust
Petiole length (mm) 5.0 25.0 45.0 65.0
0
50 Glasshouse
(173)
(121)
50
50
0 Field
Glasshouse
Final height (mm) 300.0 600.0 900.0
Glasshouse
50 Glasshouse
Field
50
0
0
50
Glasshouse
Field First Internode length (mm) 2.0 6.0 12.0 18.0
50
0 Field
50
50 Field
Hypocotyl length (mm) 15.0 25.0 35.0 45.0
Hypocotyl length (mm) 3.0 6.0 9.0 12.0
50
50
(40)
Glasshouse
Leaf length (mm) 10.0 40.0 70.0 110.0
50
Days to bolting 16.0 22.0 28.0 34.0
Restricted maximum likelihood estimates of the among-RIL variance component (VL), the residual variance component (VR), and the broadsense heritability (H2), calculated as VL/(VP); VP is the total phenotypic variance component.
Field
Fig. 1 Frequency plots of Brassica rapa recombinant inbred lines (RILs) best-linear unbiased predictors (BLUPs) in the field, noncompetitive and glasshouse, full-sun environments. Means of the RILs (solid, arrow) and the parental phenotypes for IMB211 (dotted, diamond) and R500 (dashed, oval) are shown. Ranges for time to bolting and hypocotyl length did not overlap for the field and glasshouse; these traits are thus shown on different scales for each environment. R500 means were outside the histogram for time to bolting and leaf length; values are shown. Petiole length was not recorded for IMB211 in the field.
H2 (VL/VP)
VR
VL
50
50
0 Field
Glasshouse
0.0
37.5
Days to bolting 39.5 41.5 43.5
Recombinant inbred lines differed significantly (P < 0.001) for the average expression of each trait within the field, noncompetitive and glasshouse, full-sun environments. The RIL mean was intermediate to both parental means for all traits except days to bolting in the field; we also observed transgressive segregation for most traits, in that some RILs exhibited trait values outside the range of the parental genotypes (Fig. 1). Broad-sense heritabilities (H2) were greater in the glasshouse, full-sun treatment (range, 0.30–0.71) than in the field, noncompetitive setting (range, 0.10–0.41) for all traits (Table 1). Across-environment genotypic correlations were positive for all traits (Table 2). Estimates of rGE differed significantly (P < 0.01 for RIL effect) or marginally significantly (P = 0.07 for RIL effect for leaf length) from zero for all traits except seedling
50
50
0 Field
Glasshouse
New Phytologist (2014) 201: 657–669 www.newphytologist.com
New Phytologist
662 Research
Table 2 Genotypic correlations among traits in Brassica rapa recombinant inbred lines in the field, noncompetitive (above diagonal) and glasshouse, full-sun (below diagonal) environments
Time to bolting Hypocotyl length Petiole length Leaf length Seedling internode length Final height
Time to bolting
Hypocotyl length
Petiole length
Leaf length
Seedling internode length
Final height
0.56*** (0.37)**** 0.52**** 0.45**** (0.29) *** 0.57****
(0.20)* 0.47**** (0.08) 0.06 0.72**** 0.17*
0.16 0.01 0.43** 0.85**** (0.15) 0.51****
0.07 0.05 0.68**** 0.61^ 0.04 0.61****
(0.03) 0.10 0.29** 0.23** 0.15 0.24**
0.06 0.14 0.61**** 0.79**** 0.50**** 0.64****
Cross-environmental correlations for each trait are on the diagonal (italicized). Pearson correlation coefficients are shown; parentheses indicate negative correlations. Significance is as follows: ^
