Computational protocol: Epistatic determinism of durum wheat resistance to the wheat spindle streak mosaic virus

Similar protocols

Protocol publication

[…] Two RIL (F6) populations were genotyped in 2015 and used for QTL detection based on the 2012 and 2015 data. We used locus targeted genotyping according to the protocol of Holtz et al. (). Briefly, two sets of 120 bp baits were designed to target 6240 and 10,027 SNPs previously detected in the RNAseq data of Dic2, Lloyd and Silur genitors. RIL DNA was extracted in 2015 and captured according to Rohland and Reich (). Compared to Holtz et al. (), blocking oligos were added to limit the capture of microsatellite-like sequences (Online Resource 1). Captured DNA was sequenced with two runs of HiSeq 3000 (150 bp paired end reads). Reads were cleaned and mapped on a durum wheat reference transcriptome (DWr) (Ranwez et al. ; David et al. ) using cutAdapt (Martin ) and bwa-mem (Li and Durbin ).Putative chromosomal assignment and physical positions of the DWr contigs were obtained by blasting them on the bread wheat chromosome survey sequence for cv. Chinese Spring (BWr) generated by IWGSC (Ensembl database release 28), (Mayer et al. ; Chapman et al. ). Genotypes were called using Reads2snp (Galtier et al. ) and SNPs were filtered according to the following criteria: (1) inbreeding coefficients, Fis, above 0.8 corresponding to a low probability of being heterozygotic, as 1.5% heterozygosity is expected on average after six successive selfing generations, (2) at least 100 RILs genotyped for any given SNP, and (3) balanced frequencies with a minimum expected heterozygosity (Nei ) of 0.34 so as to avoid strong segregation distortion, which is undesirable for genetic map building.Four SSR markers known to be on the distal part of chromosome 7B were also used to genotype the DS RILs: Xbarc1068, Xbarc323, Xgwm400 and Xgwm46 (http://wheat.pw.usda.gov/). These four markers are linked to SBCMV resistance (Maccaferri et al. ). They were not used for the linkage map construction, but were positioned on it afterward. SNPs showing the highest linkage disequilibrium with those SSR markers were used to position the SSRs on our consensus genetic map. This allowed us to test whether or not WSSMV and SBCMV resistance genes were collocated. [...] SNPs from the DS population (DS-SNPs) and the DL population (DL-SNPs) were used to build two individual maps (DS map and DL-map). The DS map construction was described in (Holtz et al. ), with a focus on the capture technology. The DL-map was constructed using the same procedure that is briefly described hereafter. Carthagene (de Givry et al. ) was used to assemble initial linkage groups (LGs) using LOD score thresholds of 7 and 8 for DS and DL, respectively, and maximum two-point distances of 0.14 and 0.1 for DS and DL, respectively. As the SNPs were already assigned to a BWr chromosome [best blast procedure (Holtz et al. )], each LG was then assigned to the most frequent carrier chromosome of its SNPs. Markers on LGs assigned to the same chromosome were pooled. Orders and distances between adjacent markers within chromosomes were finally determined using the build, annealing, greedy and flips algorithms implemented in Carthagene.Then common markers between the DS- and DL-maps were used to build a consensus map containing all the markers using the Carthagene dsmergen function. Genetic maps were characterized and compared using the genetic map comparator web application (Holtz et al. ). Once the marker positions were set, missing data were attributed using the CallParentAllelesPlugin of Tassel (Glaubitz et al. ). Finally, we estimated the number of recombination events accumulated during the fixation of each RIL by counting the number of switches between stretches with successive parental allelic status. [...] We used the QTLRel program (Cheng et al. ), implemented in R, for QTL analysis of the two DS and DL datasets. Sister lines (similarity >80%) were detected among our RILs after genotyping (27 in DS and 13 in DL), probably resulting from confusion during the fixation process in nurseries. Although these lines were removed from the map construction, we considered their phenotypes as worthy for increasing the QTL detection power as long as the line relatedness was explicitly declared. QTLRel implements mixed models allowing a random polygenic effect using different kinship matrices, while taking the inherent relatedness among individuals into account (Cheng et al. ). We used the GenMatrix option of QTLRel, which uses the simple kinship coefficient matrix in the model and estimates it directly from the marker data.We performed a single marker analysis on BLUP values to compute a LOD score per marker. QTL detection was first performed independently for each population and each year. A joint-QTL analysis was then performed using both populations on markers found to be polymorphic in both DS and DL populations, but all markers were used to compute the kinship correlation matrix. Finally, a last analysis was computed, taking together both years and populations, to evaluate QTL × year interaction. The LOD threshold for declaring a QTL as significant for a trait in an experiment was obtained by permutations of genotypes relative to phenotypes. We used the value of the 95% maximum LOD score values obtained from 1000 independent permutations as the significance threshold (Churchill and Doerge ). Since this threshold was always between 3.24 and 3.61, a conservative LOD threshold of 3.61 was used for all traits. QTL confidence interval regions (in cM) were defined as the ±1.5 LOD-interval around the peak LOD values of each QTL (Mangin et al. ). Allelic effects were defined as half the difference between the BLUP means of favorable and unfavorable homozygous genotypes.All R scripts and data used for those statistical and QTL analyses are available in Online Resource 4 and on Github (https://github.com/holtzy) for reproducibility. […]

Pipeline specifications