Predictions of heading date in bread wheat (Triticum aestivum L.) using QTL-based parameters of an ecophysiological model

Abstract

Prediction of wheat phenology facilitates the selection of cultivars with specific adaptations to a particular environment. However, while QTL analysis for heading date can identify major genes controlling phenology, the results are limited to the environments and genotypes tested. Moreover, while ecophysiological models allow accurate predictions in new environments, they may require substantial phenotypic data to parameterize each genotype. Also, the model parameters are rarely related to all underlying genes, and all the possible allelic combinations that could be obtained by breeding cannot be tested with models. In this study, a QTL-based model is proposed to predict heading date in bread wheat (Triticum aestivum L.). Two parameters of an ecophysiological model (V sat and P base , representing genotype vernalization requirements and photoperiod sensitivity, respectively) were optimized for 210 genotypes grown in 10 contrasting location × sowing date combinations. Multiple linear regression models predicting V sat and P base with 11 and 12 associated genetic markers accounted for 71 and 68% of the variance of these parameters, respectively. QTL-based V sat and P base estimates were able to predict heading date of an independent validation data set (88 genotypes in six location × sowing date combinations) with a root mean square error of prediction of 5 to 8.6 days, explaining 48 to 63% of the variation for heading date. The QTL-based model proposed in this study may be used for agronomic purposes and to assist breeders in suggesting locally adapted ideotypes for wheat phenology.

Keywords: Association genetics; ecophysiological model; gene-based modelling; heading date; phenology; wheat..

Fig. 1.

Sensitivity analysis of a modified version of the Weir et al. (1984) phenological model. Standardized regression coefficients showing the percentage of variation of heading date explained by each model parameter (V _sat, P _base, and TT _emhe) were calculated in each autumn- and spring-sown experiment used to optimize the model for the genotypes of the calibration data set. This figure is available in colour at JXB online.

Fig. 2.

Distributions of the V _sat (a, c), P _base (b, d) and TT _emhe (e) parameters of a modified version of the Weir et al. (1984) phenological model optimized for the 210 genotypes of a wheat-association genetics panel when two (V _sat and P _base; a, b) or three (V _sat, P _base, and TT _emh; c, d, and e) parameters were optimized. This figure is available in colour at JXB online.

Fig. 3.

Relationship between observed and predicted heading dates obtained with two optimized parameters (2p strategy) of a modified version of the Weir et al. (1984) phenological model for a wheat calibration data set grown in 10 location × sowing date combinations. Autumn- and spring-sown experiments are shown with closed symbols and stars, respectively. Winter genotypes headed only in autumn-sown experiments while spring genotypes headed in autumn- and spring-sown experiments. Symbols for autumn-sown experiments are filled in white and grey for winter and spring genotypes, respectively. Linear regression (solid) and bisecting (dashed) lines are shown. The number of data points (n), the percentage of variance explained (R²) and RMSEP are indicated.

Fig. 4.

Relationships between optimized and QTL-based predicted V _sat (a) and P _base (b) parameters of a modified version of the Weir et al. (1984) phenological model for the 210 genotypes of the calibration data set. Solid lines are regression lines and R² is the percentage of variance explained.

Fig. 5.

Relationships between observed heading dates and heading dates predicted with two optimized (a) or two QTL-based (b) parameters (2p strategy) for the calibration data set across the autumn-sown experiments using a modified version of the Weir et al. (1984) phenological model. Symbols are filled in white and grey for winter and spring genotypes, respectively. Linear regression (solid) and bisecting (dashed) lines are shown. The number of data points (n), the percentage of variance explained (R²), and RMSEP are indicated.

Fig. 6.

Relationships between observed heading dates and heading dates predicted with two QTL-based parameters (2p strategy) using a modified version of the Weir et al. (1984) phenological model for the validation data set comprising 88 independent wheat genotypes grown in six independent location × sowing date combinations. Linear regression (solid) and bisecting (dashed) lines are shown. The number of datapoints (n), the percentage of variance explained (R²), and the RMSEP are indicated.