Genetic Analysis Of Phenotypic Traits In Brassica Napus Based On High-throughput Phenotyping Platform

Brassica napus L.(B.napus)is an important oil and feed crop.In traditional research,phenotypic traits were surveyed manually using measurements tools with low efficiency and poorly reproducible.In recent years,the genetic mechanism of dynamic changes of plant phenotypes at different growth stages can be analyzed using optical imaging equipment.In this study,to revel the genetic mechanism of the dynamic growth of B.napus,an introgression population(89 lines),constructed between Zhongyou 821(ZY821)and No.2127,were investigated at different growth satges using the crop phenotyping platform of Huazhong Agricultural University.Combining the genotyping-by-sequencing,QTLs were identified to reveal the genetic basis of phenotypic traits in oilseed rape.The main results are as follows:1.In 2015 and 2016,these 89 lines were planted with 5 random replicates in the greenhouse of the phenotyping platform.From the end of November to the end of February of the next year,plant photos were taken from the side and top-view for each plant in a total of 12 periods.In each time point,23 side-view(SV)and 18 top-view(TV)phenotypic traits were extracted from the pictures.In the SV phenotypic traits,6 traits(M_TEX_SV,SE_TEX_SV,S_TEX_SV,MU3_TEX_SV,U_TEX_SV and E_TEX_SV)reflect plant texture,11 traits(GPA_SV,TPA_SV,H_SV,W_SV,HWR_SV,FDNIC_SV,FDIC_SV,R_SV,PAR_SV,HA_SV and AC_SV)reflect plant morphological and six traits(PC1_SV,PC2_SV,PC3_SV,PC4_SV,PC5_SV and PC6_SV)reflect plant compactness.Among the TV phenotypic traits,6 traits(M_TEX_TV,SE_TEX_TV,S_TEX_TV,MU3_TEX_TV,U_TEX_TV and E_TEX_TV)reflect plant texture,11 traits(GPA_TV,TPA_TV,H_TV,W_TV,HWR_TV,FDNIC_TV,FDIC_TV,R_TV,PAR_TV,HA_TV and AC_TV)reflect plant morphological and one trait(GCV_TV)reflects the degree of greenness of the leaves.2.9 lines and ZY821 were planted with 12 replicates in greenhouse.In each time point,41 phenotypic traits were measured by phenotyping platform.Then,plant height,fresh wight(FW)and dry weight(DW)were measured by manual.Taking plant height as example,the correlation of plant height measured by phenotyping platform and manual was highly significant(R~2>0.88)with small deviation(mean absolute the percentage error,MAPE<12%),indicating that the platform is reliable in measuring the phenotypic traits.The genetic basis of fresh/dry weight(FW/DW)of rapeseed was analyzed nondestructively using predict model.Firstly,stepwise regression analysis was performed between 41 phenotypic traits and FW(DW)using 120 plants.The results showed that the total projected area in side and top view(TPA_SV and TPA_TV)were significantly correlated with biomass.Secondly,predict models were constructed between TPA_SV,TPA_TV and FW(DW),including 4 linear models,2 quadratic models,2 exponential models and 3 power models.Finally,based on the adjusted coefficient(R~2)and the mean absolute percentage error(MAPE)between the model predicated and manual measured for FW(DW),the model 9(ln(y)=a+b×ln(x),x is TPA_SV,y is FW or DW,a and b are constants)is optimul model to predict FW and DW.And the 10×cross-validation shows that this model has higher prediction accuracy.Using model 9 in combination with TPA_SV for non-destructive testing of introgression populations,the predicted values of FW and DW for 89 lines can be calculated.Combining with genotypes,multiple QTLs were detected for FW and DW of rapeseed during 12 periods.3.Genetic analysis was performed for 41 phenotypic traits:(1)the heritability of all phenotypic traits showed a dynamic changes in 12 periods.And the heritability of most phenotypic traits reached more than 0.6 in multiple periods.(2)A significant negative correlation was observed between U_TEX_SV,U_TEX_TV and most phenotypic traits.The correlations of AC_SV,AC_TV,HWR_SV,HWR_TV,PAR_SV,PAR_TV,R_SV,R_TV,PC1_SV,PC2_SV,PC3_SV,PC4_SV,PC5_SV and PC6_SV with other phenotypic traits were not significant.A significant or highly significant positive correlation were observed among the remaining 25 phenotypic traits.(3)326 and 581 QTLs were identified for 41 phenotypic traits in 2015 and 2016,respectively.The character of QTL distribution are as follows:(a)all of phenotypic traits detect multiple QTLs.(b)in some single time point,multiple traits identify the same QTLs.(c)in some continuous time points,QTLs for some traits are mapped the same location.4.Stepwise regression analysis was performed using the yield per plant of 89 lines and phenotypic traits measured in 12 periods.We found 4 side-view morphological traits at the1st,11th and 12th growth stages(FDIC_SV-1,PAR_SV-1,FDNIC_SV-11 and H_SV-12)were positively correlated with the yield per plant.And the yield per plant predicted by these four traits was significantly correlated with the manually measurement(R~2=0.63)with small deviation(MAPE=5.65%).The result shows that these four traits can be used to predict plant yield.