3.6 Conclusion
4.3.3.3 Other traits
Thousand kernel weight (TKW), protein content in the defatted meal (Prot.idM), glucosinolate content in defatted meal (GSLidM) and the correction of oil content considering erucic acid content were determined according to section 3.3.3.
4.3.4 Statistical analysis
PLABSTAT software version 3A (Utz 2011) was used to perform analysis of variance (ANOVA) applying the following general model:
𝑌𝑖𝑗 = 𝜇 + 𝑔𝑖 + 𝑒𝑗+ 𝑟𝑗𝑘 + 𝑔𝑒𝑖𝑗+ ε𝑖𝑗𝑘
where Yij is the trait value of genotype i in environment j in replication k; µ is the general mean;
gi is the effect of ith genotype, ej is the effect of jth environment, rjk is the effect of replicate k in the environment j; geij is the interaction between ith genotype and jth environment; and εijk is the within environment error associated with genotype i, environment j and replicate k.
Genotypes were considered fixed in the analysis, whereas environment and replicate were treated as random variables. Years were treated as environments. The data were tested for outliers by a modification of the Anscombe and Tukey method (1963) based on the detection of extreme residuals. After examining the list of detected outliers the measured values of the outliers with highest standardized residual were checked for errors and the ANOVA was repeated considering missing values for extreme outliers. The adjusted results were used in the subsequent analyses.
Heritability (h2) was calculated as:
ℎ2= σ𝑔2
σ𝑔2 + σ𝑔𝑒2
𝐸 + σ𝜀2 𝐸𝑅
where 2g, 2ge and 2ε are variance components for g, e and ε, respectively. E and R refer to number of environments and number of replicates (Hill and Weir 1988). Spearman’s rank correlation coefficients between traits mean values of the genotypes across the environments were calculated using PLABSTAT’s BASIC command.
4.3.5 QTL mapping
QTL detection was performed as described in section 3.3.7.
4.3.6 Physical mapping
Physical mapping was conducted according to section 3.3.8.
4.4 Results
4.4.1 Phenotypic analysis
Variance analysis revealed highly significant effects of the genotypes for all traits investigated, except for flowering period (Table 4.1). In contrast, a highly significant effect of the environment was observed only for FP and for end of flowering. Glucosinolate and erucic acid content also showed significant effects of the environment. Genotype x environment interactions were significant for all traits except for protein content in the defatted meal. Effects of the genotypes showed high values for glucosinolate and erucic acid contents while highest effects of the environment were observed for end of flowering and flowering period. The highest residual error was detected for plant height. Heritability of flowering period was low with only 0.1 and heritability of end of flowering was moderate (0.55), while all other traits showed high heritabilities ranging from 0.70 for protein content to 0.99 for erucic acid content.
Table 4.1: Components of variance and heritabilities for contents of seed oil (%), protein (%), protein in defatted meal (Prot.idM in %), glucosinolates (GSL in μmol/g), glucosinolates in defatted meal (GSLidM in μmol/g), erucic acids (determined by NIRS; %) and begin of flowering (BOF), end of flowering (EOF), flowering period (FP) and plant height at end of flowering (PH_EOF in cm) in the SGEDH population
Trait Variance components Heritability
σ2g σ2e σ2ge σ2ε h²
DF 211 1 210 413
Oil 3.83 ** 0.04 0.16 ** 0.73 0.94
Protein 0.43 ** -0.02 0.07 * 0.61 0.70
Prot.idM 2.31 ** -0.02 0.08 1.11 0.88
GSL 211.64 ** 16.50 * 10.78 ** 14.81 0.96
GSLidM 689.14 ** 60.79 * 6.05 ** 3.52 0.96
NIRS22:1 266.26 ** 2.53 * 2.88 ** 5.72 0.99
BOFa 4.48 ** 0.00 0.73 ** 1.36 0.86
EOFa 1.93 ** 33.20 ** 1.76 ** 2.72 0.55
FP 0.20 35.10 ** 2.05 ** 2.90 0.10
PH_EOF 89.50 ** -0.03 8.23 * 64.39 0.82
σ2g = genetic variance; σ2e = variance of the environment; σ2ge = variance of genotype x environment interaction; σ2ε = residual error; DF = degrees of freedom; *, ** denotes significance at P < 5% and 1%
a days counted from 1st of January
Transgressive segregation was observed for all traits (Table 4.2). Oil content ranged from 39.4 to 49.8%, with a mean value of 44.5%. For contents of oil, glucosinolates, glucosinolates in defatted meal and erucic acid as well as plant height SGDH14 showed higher values compared to Express617, while for protein, protein in defatted meal, end of flowering and flowering period Express617 exceeded SGDH14. The parental lines showed similar values for begin of flowering, differing only 0.2 days.
Table 4.2: Minimum, maximum and mean values for contents of seed oil (%), protein (%), protein in
LSD 5% = least significant difference at P < 5%; ** denotes significance at P < 1%
Oil content showed highly significant positive correlation to erucic acid (0.57) and protein in defatted meal (0.54) as well as to plant height (0.37) and end of flowering (0.28) (Table 4.3). A highly significant negative correlation was found for oil content and protein content. Protein content was positively correlated to protein content in defatted meal, while protein content in defatted meal was positively correlated to erucic acid content. A close correlation was also observed between begin of flowering and end of flowering as well as plant height, but a negative correlation was found with the flowering period. End of flowering and plant height showed highly significant positive correlation.
Table 4.3: Spearman’s rank correlation for seed oil content and other quality traits
Traits Oil Protein Prot.idM GSL GSLidM C22:1 BOF EOF FP
The linear relation between oil content and erucic acid content showed a segregation of the DH population for erucic acid (Figure 4.1a). One group of low erucic acid content comprising 70 genotypes showed erucic acid contents ranging from -5.8 to 2.3%, and oil contents of 39.4 to 45.4%. Medium and high erucic acid genotypes did not show a clear separation. According to the grouping of the European trials (section 3.4.1) the group of genotypes (84) with medium erucic acid content showed erucic acid contents between 6.8 to 29.3%, and had oil contents between 42.7 and 48.1%. The group of genotypes (58) with high erucic acid content showed erucic acid contents between 30.2 to 46.1% erucic acid in which oil content ranged from 41.9 to 49.8%.
Within the group of low erucic acid genotypes SGEDH172 was detected as the genotype with highest oil content (45.4%). Regression correction of oil contents (Figure 4.1b) nearly eliminated the effect of erucic acid on oil content, reducing the coefficient of determination to 0.0000002. Regression correction calculated the theoretical erucic acid free oil content of SGDH14 to 44.1%. Comparing the oil contents of all genotypes of the SGEDH population applying the regression correction DH line 210 (SGEDH210) and 145 (SGEDH145) were identified as genotypes with highest oil contents, with 45.9% and 45.8% oil, respectively.
Figure 4.1: Correlation between corrected NIRS predicted seed erucic acid content and (a) NIRS predicted oil content, and (b) NIRS predicted oil content corrected for the influence of erucic acid by regression information (regression corrected oil content = oil content – (0.1 * erucic acid content)) in the SGEDH population; NIRS = near infrared reflectance spectroscopy; DM = dry matter
y = 0.1x + 42.75
regression corrected NIRS oil content in % (91% DM)
adjusted NIRS erucic acid content in % (b)
4.4.2 QTL mapping
4.4.2.1 QTL for oil content using original data
QTL mapping identified six QTL for oil content on the linkage groups A06, A07, A08, A10, C03 and C05. These QTL individually explained between 3.3 and 46.5% of the phenotypic variance and collectively accounted for 75.8% of the total phenotypic variance. The major QTL C_Oil-3 on A08 and C_Oil-5 on C03 explained 46.5 and 29.4% of the phenotypic variance, respectively.
Except C_Oil-4 on linkage group A10, all QTL showed positive additive effects, indicating that the alleles increasing oil content are derived from SGDH14. On linkage group A06 C_Oil-1 overlapped with C_PH_EOF-2 both showing a positive additive effect according to their positive correlation. On linkage group A08 major QTL C_NIRS22:1-2 and C_Prot.idM-2 overlapped with C_Oil-3. All three QTL showed positive additive effects, indicating that alleles derived from SGDH14 increased oil content by 1.06%, erucic acid content by 8.7% and protein content in the defatted meal by 0.91%, respectively. C_Oil-5 on C03, the second major QTL for oil content was co-located with the major QTL C_NIRS22:1-3 for erucic acid content, both showing a positive additive effect, indicating alleles derived from SGDH14 were increasing oil content by 0.73% and erucic acid by 7.58%, respectively. C_Oil-4 located on A10 showed an individual position.
4.4.2.2 QTL for oil content corrected for the effect of erucic acid content
Three QTL were detected for regression corrected and for conditioned oil content located on linkage groups A06, A07 and C05 with identical positions and almost identical confidence intervals. Individual QTL explained 6.7% (A06), 12.7% (A07) and 17.2% (C05) of the phenotypic variance and all three QTL accounted for 33.9% of the total phenotypic variance. All three QTL showed positive additive effects, indicating that alleles derived from SGDH14 were increasing corrected oil contents by 0.32%, 0.43% and 0.46%, respectively. QTL for corrected oil contents all overlapped with QTL for oil content on respective linkage groups.
4.4.2.3 QTL for seed protein content
Two QTL for protein content were detected on the linkage groups A09 and C08 explaining 9.6%
and 7.2% of the phenotypic variance respectively and together accounting for 16.5% of the total phenotypic variance. Both QTL showed negative additive effects, indicating that alleles derived from Express617 were increasing protein content.
4.4.2.4 QTL for protein content in defatted meal
The five QTL for seed protein content in the defatted meal were detected on linkage groups A07, A08, C01, C03 and C04. Individual QTL explained between 0.7 and 39.8% of the phenotypic variance and collectively explained 56% of the total phenotypic variance. 4.9% of the total phenotypic variance (60.9%) was explained by epistatic interactions. C_Prot.idM-3 and C_Prot.idM-5 showed negative additive effects, indicating that alleles increasing protein content in the defatted meal were derived from Express617. While the other three QTL showed positive additive effects including the major QTL C_Prot.idM-2, which increased protein content in defatted meal by 0.91%.
4.4.2.5 QTL for glucosinolate content
Five QTL for glucosinolate content were detected on linkage groups A09, C02 (2), C07 and C09, each explaining between 0 and 58% of the phenotypic variance. All QTL together accounted for 81% of the total phenotypic variance. QTL C_GSL-5 was identified as the major QTL explaining 58% of the phenotypic variance. The additive effect of C_GSL-5 was positive, indicating that an allele derived from SGDH14 was increasing glucosinolate content by 11.4µmol/g. QTL C_GSL-1, C_GSL-2 and C_GSL-4 also showed positive additive effects. Only C_GSL-3 on C02 explaining 0%
of the phenotypic variance showed a negative additive effect.
4.4.2.6 QTL for glucosinolate content in defatted meal
Seven QTL for glucosinolate content were detected on linkage groups A04, A09, C02 (2), C07 and C09 (2) each explaining between 0 and 55.6% of the phenotypic variance. All QTL together accounted for 81% of the total phenotypic variance. Additional 7.3% of the total phenotypic variance was explained by epistatic interactions. Among the seven QTL, C_GSLidM-6 and C_GSLidM-7 were identified as major QTL explaining 55.2% and 55.6% of the phenotypic variance, respectively. The additive effect of C_GSLidM-1 and C_GSLidM-4 were negative, indicating that an allele derived from Express617 was increasing glucosinolate content by 2.4µmol/g and 3.6µmol/g, respectively. Except C_GSLidM-1 on A04 and C_GSLidM-6 on C09, all other QTL for glucosinolate content in defatted meal were co-located with QTL for glucosinolate content.