| Most forest trees have the characteristics of long generation times,complex genetic structure,different types of marker segregation and unknown linkage phases.In the last thirty years,statistical analysis of QTL mapping in forest trees has been greatly developed,but there are still a lot of problems,especially in the case involved multivariate trait data where no relevant software can be used directly.In this study,we focused on the multivariate trait data for QTL mapping in forest trees,developed the statistical models and wrote the related software.We performed QTL mapping with the new developed methods for tree heights over multiple time points in a hybrid population of Populus.We first applied the interval mapping method into F1 hybrid population in forest trees for multivariate trait data by considering various possible segregation types of molecular markers and their linkage information.An R package called mvqtlmap was developed to implement this algorithm and it can be downloaded at http://www.bioseqdata.com/mvqtlmap/mvqtlmap.htm.Meanwhile,we impelemented the traditional composite interval mapping（CIM）to multivariate data and developed the corresponding software,mvqtlcim.This method not only incorporated the various segregations and linkage phases of markers,but also applied Takeuchi’s information criterion（TIC）to discriminate the QTL segregation type among several possible alternatives.The web site of this software is at https://github.com/tongchf/mvqtlcim.The software package also allows a lot of parameters to be selected and parallel computations for permutaion tests.The features of the two new softwares were verified with the Poplar data analyses and a large number of Monte Carlo simulations.In a hybrid F1 population of P.deltoides and P.simonii,we performed interval QTL mapping for tree heights over 6 different time points with the R package mvqtlmap developed here.As a result,4 QTLs affecting tree heights were located on the female linkage of P.deltoides,and 6QTL were found on the male linkage map of P.simonii.These QTLs distributed on chromosomes1,5,7,9,11 and 19,and each QTL could explain 0.8%to 6.7%of the phenotypic variance.With the same phenotype data,we performed CIM QTL mapping using the new developed software mvqtcim.Consequently,12 QTLs were detected for tree height over 6 time points.Of these QTLs,10 were located on the female linkage map of P.deltoides,distributing on chromosomes 1,2,5,9 and 14,and 2 on the male linkage map of P.simonii,distributing on chromosomes 7 and 9.Each of them could explain 1.1%to 27.4%of the phenotypic variance.We searched the candidate genes of the 12 QTLs idedtified with CIM method in the coding region of each QTL’s physical interval and the gene annotation database of P.trichocarpa.The coding sequences（CDS）of those genes associated with each QTL were re-annotated by blasting first and then mapping on Gene Ontology（GO）terms.We found that the genomic region covering a QTL has an average length of 801 kb and contained 7–247 genes,of which 79%have19.7 blast hits and 5.0 GO terms received on average.Two candidate genes in the Q1D14 and one gene of QD5 are related to the shoot formation or development.We also found candidate genes of embryo or root development in flanking regions of Q1D14,QD5,QD9,QS7 and QS9,and genes related to photosynthesis are located in regions of Q1D2,Q2D2,Q1D14,QD5 and QS9.In order to compare with other methods for QTL mapping,we conducted QTL analysis with our Populus real marker datasets each from one parental linkage map using the popular LASSO method.As a result,a total of 12 SNPs were identified to be associated with the tree height,exactly half of which came from each dataset.Among these SNPs,three were identified consistently by both CIM and LASSO.The results from the same real data analyses display some differences in using different QTL mapping methods.Therefore,more efforts should be paid to verifying the detected QTLs as well as to improving the QTL mapping methods continuously.The software developed in this study provide a powerful tool for mapping QTLs with multivariate trait data,and thus will accelerate the molecular breeding prograns especially in forest trees. |
